100% found this document useful (2 votes)

2K views1,787 pages

Pandas PDF

Uploaded by

ArjunaPamungkas

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

100% found this document useful (2 votes)

2K views1,787 pages

Pandas PDF

Uploaded by

ArjunaPamungkas

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 1787

pandas: powerful Python data analysis

toolkit
Release 0.17.0

Wes McKinney & PyData Development Team

October 09, 2015

CONTENTS

Whats New
1.1 v0.17.0 (October 9, 2015) . . . . . . . . . . . . . . . .
1.2 v0.16.2 (June 12, 2015) . . . . . . . . . . . . . . . . .
1.3 v0.16.1 (May 11, 2015) . . . . . . . . . . . . . . . . .
1.4 v0.16.0 (March 22, 2015) . . . . . . . . . . . . . . . .
1.5 v0.15.2 (December 12, 2014) . . . . . . . . . . . . . .
1.6 v0.15.1 (November 9, 2014) . . . . . . . . . . . . . . .
1.7 v0.15.0 (October 18, 2014) . . . . . . . . . . . . . . . .
1.8 v0.14.1 (July 11, 2014) . . . . . . . . . . . . . . . . . .
1.9 v0.14.0 (May 31 , 2014) . . . . . . . . . . . . . . . . .
1.10 v0.13.1 (February 3, 2014) . . . . . . . . . . . . . . . .
1.11 v0.13.0 (January 3, 2014) . . . . . . . . . . . . . . . .
1.12 v0.12.0 (July 24, 2013) . . . . . . . . . . . . . . . . . .
1.13 v0.11.0 (April 22, 2013) . . . . . . . . . . . . . . . . .
1.14 v0.10.1 (January 22, 2013) . . . . . . . . . . . . . . . .
1.15 v0.10.0 (December 17, 2012) . . . . . . . . . . . . . .
1.16 v0.9.1 (November 14, 2012) . . . . . . . . . . . . . . .
1.17 v0.9.0 (October 7, 2012) . . . . . . . . . . . . . . . . .
1.18 v0.8.1 (July 22, 2012) . . . . . . . . . . . . . . . . . .
1.19 v0.8.0 (June 29, 2012) . . . . . . . . . . . . . . . . . .
1.20 v.0.7.3 (April 12, 2012) . . . . . . . . . . . . . . . . .
1.21 v.0.7.2 (March 16, 2012) . . . . . . . . . . . . . . . . .
1.22 v.0.7.1 (February 29, 2012) . . . . . . . . . . . . . . .
1.23 v.0.7.0 (February 9, 2012) . . . . . . . . . . . . . . . .
1.24 v.0.6.1 (December 13, 2011) . . . . . . . . . . . . . . .
1.25 v.0.6.0 (November 25, 2011) . . . . . . . . . . . . . . .
1.26 v.0.5.0 (October 24, 2011) . . . . . . . . . . . . . . . .
1.27 v.0.4.3 through v0.4.1 (September 25 - October 9, 2011)

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

3
3
28
32
42
58
63
69
96
102
129
136
159
170
179
185
196
200
202
202
208
211
212
212
217
218
220
221

Installation
2.1 Python version support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Installing pandas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3 Dependencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

223
223
223
226

Contributing to pandas
3.1 Where to start? . . . . . . . . . . . .
3.2 Bug Reports/Enhancement Requests .
3.3 Working with the code . . . . . . . .
3.4 Contributing to the documentation . .
3.5 Contributing to the code base . . . .
3.6 Contributing your changes to pandas

229
229
230
230
233
235
238

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

Frequently Asked Questions (FAQ)

241
4.1 DataFrame memory usage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
4.2 Byte-Ordering Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
4.3 Visualizing Data in Qt applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243

Package overview
5.1 Data structures at a glance . . .
5.2 Mutability and copying of data .
5.3 Getting Support . . . . . . . .
5.4 Credits . . . . . . . . . . . . .
5.5 Development Team . . . . . . .
5.6 License . . . . . . . . . . . . .

.
.
.
.
.
.

245
245
246
246
246
246
246

10 Minutes to pandas
6.1 Object Creation . . .
6.2 Viewing Data . . . .
6.3 Selection . . . . . .
6.4 Missing Data . . . .
6.5 Operations . . . . .
6.6 Merge . . . . . . . .
6.7 Grouping . . . . . .
6.8 Reshaping . . . . .
6.9 Time Series . . . . .
6.10 Categoricals . . . .
6.11 Plotting . . . . . . .
6.12 Getting Data In/Out
6.13 Gotchas . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.

249
249
251
252
257
257
260
262
263
265
266
268
269
271

Tutorials
7.1 Internal Guides . . . . . . . . . . . . . . . . . .
7.2 pandas Cookbook . . . . . . . . . . . . . . . .
7.3 Lessons for New pandas Users . . . . . . . . . .
7.4 Practical data analysis with Python . . . . . . .
7.5 Excel charts with pandas, vincent and xlsxwriter
7.6 Various Tutorials . . . . . . . . . . . . . . . . .

.
.
.
.
.
.

273
273
273
274
274
274
274

Cookbook
8.1 Idioms . . . . . . . . .
8.2 Selection . . . . . . . .
8.3 MultiIndexing . . . . .
8.4 Missing Data . . . . . .
8.5 Grouping . . . . . . . .
8.6 Timeseries . . . . . . .
8.7 Merge . . . . . . . . . .
8.8 Plotting . . . . . . . . .
8.9 Data In/Out . . . . . . .
8.10 Computation . . . . . .
8.11 Timedeltas . . . . . . .
8.12 Aliasing Axis Names . .
8.13 Creating Example Data .

.
.
.
.
.
.
.
.
.
.
.
.
.

277
277
280
284
288
289
297
298
299
300
303
304
305
306

.
.
.
.
.
.
.
.
.
.
.
.
.

Intro to Data Structures

307
9.1 Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
9.2 DataFrame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
9.3 Panel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326

9.4
9.5

Panel4D (Experimental) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330

PanelND (Experimental) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332

10 Essential Basic Functionality

10.1 Head and Tail . . . . . . . . . . . .
10.2 Attributes and the raw ndarray(s) .
10.3 Accelerated operations . . . . . . .
10.4 Flexible binary operations . . . . .
10.5 Descriptive statistics . . . . . . . .
10.6 Function application . . . . . . . .
10.7 Reindexing and altering labels . . .
10.8 Iteration . . . . . . . . . . . . . . .
10.9 .dt accessor . . . . . . . . . . . . .
10.10 Vectorized string methods . . . . .
10.11 Sorting . . . . . . . . . . . . . . .
10.12 Copying . . . . . . . . . . . . . .
10.13 dtypes . . . . . . . . . . . . . . . .
10.14 Selecting columns based on dtype

.
.
.
.
.
.
.
.
.
.
.
.
.
.

335
335
336
337
337
344
352
358
366
369
372
373
377
377
384

11 Working with Text Data

11.1 Splitting and Replacing Strings
11.2 Indexing with .str . . . . . .
11.3 Extracting Substrings . . . . . .
11.4 Method Summary . . . . . . .

.
.
.
.

387
388
391
391
393

12 Options and Settings

12.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . .
12.2 Getting and Setting Options . . . . . . . . . . . . . . .
12.3 Setting Startup Options in python/ipython Environment
12.4 Frequently Used Options . . . . . . . . . . . . . . . . .
12.5 List of Options . . . . . . . . . . . . . . . . . . . . . .
12.6 Number Formatting . . . . . . . . . . . . . . . . . . .
12.7 Unicode Formatting . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.

395
395
396
397
397
403
404
404

13 Indexing and Selecting Data

13.1 Different Choices for Indexing . . . . .
13.2 Basics . . . . . . . . . . . . . . . . . .
13.3 Attribute Access . . . . . . . . . . . .
13.4 Slicing ranges . . . . . . . . . . . . .
13.5 Selection By Label . . . . . . . . . . .
13.6 Selection By Position . . . . . . . . .
13.7 Selecting Random Samples . . . . . .
13.8 Setting With Enlargement . . . . . . .
13.9 Fast scalar value getting and setting . .
13.10 Boolean indexing . . . . . . . . . . . .
13.11 Indexing with isin . . . . . . . . . . .
13.12 The where() Method and Masking .
13.13 The query() Method (Experimental)
13.14 Duplicate Data . . . . . . . . . . . . .
13.15 Dictionary-like get() method . . . .
13.16 The select() Method . . . . . . . .
13.17 The lookup() Method . . . . . . . .
13.18 Index objects . . . . . . . . . . . . . .
13.19 Set / Reset Index . . . . . . . . . . . .
13.20 Returning a view versus a copy . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

407
407
408
410
412
413
416
419
421
422
423
425
427
430
441
443
443
444
444
446
449

.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

iii

14 MultiIndex / Advanced Indexing

14.1 Hierarchical indexing (MultiIndex) . . . . .
14.2 Advanced indexing with hierarchical index .
14.3 The need for sortedness with MultiIndex
14.4 Take Methods . . . . . . . . . . . . . . . .
14.5 CategoricalIndex . . . . . . . . . . . . . . .
14.6 Float64Index . . . . . . . . . . . . . . . . .

.
.
.
.
.
.

453
453
459
468
470
471
474

15 Computational tools
15.1 Statistical functions . . . . . . . . . . . .
15.2 Moving (rolling) statistics / moments . . .
15.3 Expanding window moment functions . . .
15.4 Exponentially weighted moment functions

.
.
.
.

479
479
483
491
492

16 Working with missing data

16.1 Missing data basics . . . . . . . . . . .
16.2 Datetimes . . . . . . . . . . . . . . . .
16.3 Inserting missing data . . . . . . . . .
16.4 Calculations with missing data . . . . .
16.5 Cleaning / filling missing data . . . . .
16.6 Missing data casting rules and indexing

.
.
.
.
.
.

497
497
499
500
501
502
516

17 Group By: split-apply-combine

17.1 Splitting an object into groups .
17.2 Iterating through groups . . . .
17.3 Selecting a group . . . . . . . .
17.4 Aggregation . . . . . . . . . .
17.5 Transformation . . . . . . . . .
17.6 Filtration . . . . . . . . . . . .
17.7 Dispatching to instance methods
17.8 Flexible apply . . . . . . . .
17.9 Other useful features . . . . . .
17.10 Examples . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.

519
520
524
525
526
529
533
534
536
538
545

.
.
.
.
.
.
.
.
.
.

18 Merge, join, and concatenate

547
18.1 Concatenating objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 547
18.2 Database-style DataFrame joining/merging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 558
19 Reshaping and Pivot Tables
19.1 Reshaping by pivoting DataFrame objects
19.2 Reshaping by stacking and unstacking . .
19.3 Reshaping by Melt . . . . . . . . . . . .
19.4 Combining with stats and GroupBy . . .
19.5 Pivot tables and cross-tabulations . . . .
19.6 Tiling . . . . . . . . . . . . . . . . . . .
19.7 Computing indicator / dummy variables .
19.8 Factorizing values . . . . . . . . . . . .

.
.
.
.
.
.
.
.

573
573
574
579
580
581
585
585
588

20 Time Series / Date functionality

20.1 Overview . . . . . . . . . . . . .
20.2 Time Stamps vs. Time Spans . .
20.3 Converting to Timestamps . . . .
20.4 Generating Ranges of Timestamps
20.5 DatetimeIndex . . . . . . . . . .
20.6 DateOffset objects . . . . . . . .

.
.
.
.
.
.

589
590
590
591
594
596
602

.
.
.
.
.
.

20.7
20.8
20.9
20.10
20.11
20.12

Time series-related instance methods

Resampling . . . . . . . . . . . . . .
Time Span Representation . . . . . .
Converting between Representations
Representing out-of-bounds spans . .
Time Zone Handling . . . . . . . . .

21 Time Deltas
21.1 Parsing . . . . . . . .
21.2 Operations . . . . . .
21.3 Reductions . . . . . .
21.4 Frequency Conversion
21.5 Attributes . . . . . . .
21.6 TimedeltaIndex . . . .
21.7 Resampling . . . . . .

.
.
.
.
.
.

614
616
619
625
626
627

.
.
.
.
.
.
.

637
637
638
642
643
644
645
648

22 Categorical Data
22.1 Object Creation . . . . .
22.2 Description . . . . . . .
22.3 Working with categories
22.4 Sorting and Order . . .
22.5 Comparisons . . . . . .
22.6 Operations . . . . . . .
22.7 Data munging . . . . .
22.8 Getting Data In/Out . .
22.9 Missing Data . . . . . .
22.10 Differences to Rs factor
22.11 Gotchas . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.

649
649
652
653
656
658
660
661
665
666
667
668

23 Plotting
23.1 Basic Plotting: plot . . . . . .
23.2 Other Plots . . . . . . . . . . .
23.3 Plotting with Missing Data . . .
23.4 Plotting Tools . . . . . . . . . .
23.5 Plot Formatting . . . . . . . . .
23.6 Plotting directly with matplotlib
23.7 Trellis plotting interface . . . .

.
.
.
.
.
.
.

673
673
676
707
708
716
739
740

24 IO Tools (Text, CSV, HDF5, ...)

24.1 CSV & Text files . . . . . . . . .
24.2 JSON . . . . . . . . . . . . . . .
24.3 HTML . . . . . . . . . . . . . .
24.4 Excel files . . . . . . . . . . . .
24.5 Clipboard . . . . . . . . . . . . .
24.6 Pickling . . . . . . . . . . . . . .
24.7 msgpack (experimental) . . . . .
24.8 HDF5 (PyTables) . . . . . . . . .
24.9 SQL Queries . . . . . . . . . . .
24.10 Google BigQuery (Experimental)
24.11 Stata Format . . . . . . . . . . .
24.12 Other file formats . . . . . . . . .
24.13 SAS Format . . . . . . . . . . .
24.14 Performance Considerations . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.

755
756
780
788
795
801
802
802
805
832
841
845
847
847
848

25 Remote Data Access

851
v

25.1
25.2
25.3
25.4
25.5
25.6
25.7

Yahoo! Finance . . . . .
Yahoo! Finance Options
Google Finance . . . . .
FRED . . . . . . . . . .
Fama/French . . . . . .
World Bank . . . . . . .
Google Analytics . . . .

.
.
.
.
.
.
.

851
852
854
854
855
855
858

26 Enhancing Performance
861
26.1 Cython (Writing C extensions for pandas) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 861
26.2 Using numba . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 865
26.3 Expression Evaluation via eval() (Experimental) . . . . . . . . . . . . . . . . . . . . . . . . . . . 867
27 Sparse data structures
27.1 SparseArray . . . . . . . .
27.2 SparseList . . . . . . . . .
27.3 SparseIndex objects . . . .
27.4 Interaction with scipy.sparse

.
.
.
.

875
877
877
878
878

28 Caveats and Gotchas

28.1 Using If/Truth Statements with pandas . . . . .
28.2 NaN, Integer NA values and NA type promotions
28.3 Integer indexing . . . . . . . . . . . . . . . . .
28.4 Label-based slicing conventions . . . . . . . . .
28.5 Miscellaneous indexing gotchas . . . . . . . . .
28.6 Timestamp limitations . . . . . . . . . . . . . .
28.7 Parsing Dates from Text Files . . . . . . . . . .
28.8 Differences with NumPy . . . . . . . . . . . . .
28.9 Thread-safety . . . . . . . . . . . . . . . . . . .
28.10 HTML Table Parsing . . . . . . . . . . . . . . .
28.11 Byte-Ordering Issues . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.

883
883
884
886
886
887
889
889
890
890
890
891

29 rpy2 / R interface
29.1 Updating your code to use rpy2 functions
29.2 R interface with rpy2 . . . . . . . . . . .
29.3 Transferring R data sets into Python . . .
29.4 Converting DataFrames into R objects . .
29.5 Calling R functions with pandas objects .
29.6 High-level interface to R estimators . . .

.
.
.
.
.
.

893
893
894
894
894
895
895

30 pandas Ecosystem
30.1 Statistics and Machine Learning
30.2 Visualization . . . . . . . . . .
30.3 IDE . . . . . . . . . . . . . . .
30.4 API . . . . . . . . . . . . . . .
30.5 Domain Specific . . . . . . . .
30.6 Out-of-core . . . . . . . . . . .

.
.
.
.
.
.

897
897
897
898
899
899
899

31 Comparison with R / R libraries

31.1 Base R . . . . . . . . . . .
31.2 zoo . . . . . . . . . . . . .
31.3 xts . . . . . . . . . . . . .
31.4 plyr . . . . . . . . . . . . .
31.5 reshape / reshape2 . . . . .

.
.
.
.
.

901
901
907
907
907
908

.
.
.
.

.
.
.
.
.

.
.
.
.

.
.
.
.
.

.
.
.
.

32 Comparison with SQL

32.1 SELECT . . . .
32.2 WHERE . . . .
32.3 GROUP BY . .
32.4 JOIN . . . . . .
32.5 UNION . . . . .
32.6 UPDATE . . . .
32.7 DELETE . . . .

.
.
.
.
.
.
.

913
913
914
916
918
919
921
921

33 Comparison with SAS

33.1 Data Structures . . .
33.2 Data Input / Output .
33.3 Data Operations . .
33.4 Merging . . . . . .
33.5 Missing Data . . . .
33.6 GroupBy . . . . . .
33.7 Other Considerations

.
.
.
.
.
.
.

923
923
924
925
929
930
932
934

34 API Reference
34.1 Input/Output . . . . . .
34.2 General functions . . . .
34.3 Series . . . . . . . . . .
34.4 DataFrame . . . . . . .
34.5 Panel . . . . . . . . . .
34.6 Panel4D . . . . . . . .
34.7 Index . . . . . . . . . .
34.8 CategoricalIndex . . . .
34.9 DatetimeIndex . . . . .
34.10 TimedeltaIndex . . . . .
34.11 GroupBy . . . . . . . .
34.12 General utility functions

.
.
.
.
.
.
.
.
.
.
.
.

935
935
963
1006
1180
1373
1465
1519
1552
1579
1612
1634
1656

.
.
.
.
.
.
.

35 Internals
1669
35.1 Indexing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1669
35.2 Subclassing pandas Data Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1670
36 Release Notes
36.1 pandas 0.17.0
36.2 pandas 0.16.2
36.3 pandas 0.16.1
36.4 pandas 0.16.0
36.5 pandas 0.15.2
36.6 pandas 0.15.1
36.7 pandas 0.15.0
36.8 pandas 0.14.1
36.9 pandas 0.14.0
36.10 pandas 0.13.1
36.11 pandas 0.13.0
36.12 pandas 0.12.0
36.13 pandas 0.11.0
36.14 pandas 0.10.1
36.15 pandas 0.10.0
36.16 pandas 0.9.1
36.17 pandas 0.9.0
36.18 pandas 0.8.1

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

1675
1675
1679
1680
1682
1684
1686
1687
1689
1691
1694
1697
1711
1718
1724
1726
1731
1733
1738
vii

36.19
36.20
36.21
36.22
36.23
36.24
36.25
36.26
36.27
36.28
36.29
36.30
36.31

pandas 0.8.0
pandas 0.7.3
pandas 0.7.2
pandas 0.7.1
pandas 0.7.0
pandas 0.6.1
pandas 0.6.0
pandas 0.5.0
pandas 0.4.3
pandas 0.4.2
pandas 0.4.1
pandas 0.4.0
pandas 0.3.0

Python Module Index

viii

.
.
.
.
.
.
.
.
.
.
.
.
.

1740
1745
1746
1748
1749
1755
1757
1761
1765
1766
1767
1769
1774
1777

pandas: powerful Python data analysis toolkit, Release 0.17.0

PDF Version
Zipped HTML Date: October 09, 2015 Version: 0.17.0
Binary Installers: http://pypi.python.org/pypi/pandas
Source Repository: http://github.com/pydata/pandas
Issues & Ideas: https://github.com/pydata/pandas/issues
Q&A Support: http://stackoverflow.com/questions/tagged/pandas
Developer Mailing List: http://groups.google.com/group/pydata
pandas is a Python package providing fast, flexible, and expressive data structures designed to make working with
relational or labeled data both easy and intuitive. It aims to be the fundamental high-level building block for doing
practical, real world data analysis in Python. Additionally, it has the broader goal of becoming the most powerful
and flexible open source data analysis / manipulation tool available in any language. It is already well on its way
toward this goal.
pandas is well suited for many different kinds of data:
Tabular data with heterogeneously-typed columns, as in an SQL table or Excel spreadsheet
Ordered and unordered (not necessarily fixed-frequency) time series data.
Arbitrary matrix data (homogeneously typed or heterogeneous) with row and column labels
Any other form of observational / statistical data sets. The data actually need not be labeled at all to be placed
into a pandas data structure
The two primary data structures of pandas, Series (1-dimensional) and DataFrame (2-dimensional), handle the
vast majority of typical use cases in finance, statistics, social science, and many areas of engineering. For R users,
DataFrame provides everything that Rs data.frame provides and much more. pandas is built on top of NumPy
and is intended to integrate well within a scientific computing environment with many other 3rd party libraries.
Here are just a few of the things that pandas does well:
Easy handling of missing data (represented as NaN) in floating point as well as non-floating point data
Size mutability: columns can be inserted and deleted from DataFrame and higher dimensional objects
Automatic and explicit data alignment: objects can be explicitly aligned to a set of labels, or the user can
simply ignore the labels and let Series, DataFrame, etc. automatically align the data for you in computations
Powerful, flexible group by functionality to perform split-apply-combine operations on data sets, for both aggregating and transforming data
Make it easy to convert ragged, differently-indexed data in other Python and NumPy data structures into
DataFrame objects
Intelligent label-based slicing, fancy indexing, and subsetting of large data sets
Intuitive merging and joining data sets
Flexible reshaping and pivoting of data sets
Hierarchical labeling of axes (possible to have multiple labels per tick)
Robust IO tools for loading data from flat files (CSV and delimited), Excel files, databases, and saving / loading
data from the ultrafast HDF5 format
Time series-specific functionality: date range generation and frequency conversion, moving window statistics,
moving window linear regressions, date shifting and lagging, etc.

CONTENTS

pandas: powerful Python data analysis toolkit, Release 0.17.0

Many of these principles are here to address the shortcomings frequently experienced using other languages / scientific
research environments. For data scientists, working with data is typically divided into multiple stages: munging and
cleaning data, analyzing / modeling it, then organizing the results of the analysis into a form suitable for plotting or
tabular display. pandas is the ideal tool for all of these tasks.
Some other notes
pandas is fast. Many of the low-level algorithmic bits have been extensively tweaked in Cython code. However,
as with anything else generalization usually sacrifices performance. So if you focus on one feature for your
application you may be able to create a faster specialized tool.
pandas is a dependency of statsmodels, making it an important part of the statistical computing ecosystem in
Python.
pandas has been used extensively in production in financial applications.
Note: This documentation assumes general familiarity with NumPy. If you havent used NumPy much or at all, do
invest some time in learning about NumPy first.
See the package overview for more detail about whats in the library.

CONTENTS

CHAPTER

ONE

WHATS NEW

These are new features and improvements of note in each release.

1.1 v0.17.0 (October 9, 2015)

This is a major release from 0.16.2 and includes a small number of API changes, several new features, enhancements,
and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this
version.
Warning: pandas >= 0.17.0 will no longer support compatibility with Python version 3.2 (GH9118)
Warning: The pandas.io.data package is deprecated and will be replaced by the pandas-datareader package. This will allow the data modules to be independently updated to your pandas installation. The API for
pandas-datareader v0.1.1 is exactly the same as in pandas v0.17.0 (GH8961, GH10861).
After installing pandas-datareader, you can easily change your imports:
from pandas.io import data, wb

becomes
from pandas_datareader import data, wb

Highlights include:
Release the Global Interpreter Lock (GIL) on some cython operations, see here
Plotting methods are now available as attributes of the .plot accessor, see here
The sorting API has been revamped to remove some long-time inconsistencies, see here
Support for a datetime64[ns] with timezones as a first-class dtype, see here
The default for to_datetime will now be to raise when presented with unparseable formats, previously
this would return the original input. Also, date parse functions now return consistent results. See here
The default for dropna in HDFStore has changed to False, to store by default all rows even if they are all
NaN, see here
Datetime accessor (dt) now supports Series.dt.strftime to generate formatted strings for datetimelikes, and Series.dt.total_seconds to generate each duration of the timedelta in seconds. See here
Period and PeriodIndex can handle multiplied freq like 3D, which corresponding to 3 days span. See here
Development installed versions of pandas will now have PEP440 compliant version strings (GH9518)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Development support for benchmarking with the Air Speed Velocity library (GH8361)
Support for reading SAS xport files, see here
Documentation comparing SAS to pandas, see here
Removal of the automatic TimeSeries broadcasting, deprecated since 0.8.0, see here
Display format with plain text can optionally align with Unicode East Asian Width, see here
Compatibility with Python 3.5 (GH11097)
Compatibility with matplotlib 1.5.0 (GH11111)
Check the API Changes and deprecations before updating.
Whats new in v0.17.0
New features
Datetime with TZ
Releasing the GIL
Plot submethods
Additional methods for dt accessor
* strftime
* total_seconds
Period Frequency Enhancement
Support for SAS XPORT files
Support for Math Functions in .eval()
Changes to Excel with MultiIndex
Google BigQuery Enhancements
Display Alignment with Unicode East Asian Width
Other enhancements
Backwards incompatible API changes
Changes to sorting API
Changes to to_datetime and to_timedelta
* Error handling
* Consistent Parsing
Changes to Index Comparisons
Changes to Boolean Comparisons vs. None
HDFStore dropna behavior
Changes to display.precision option
Changes to Categorical.unique
Changes to bool passed as header in Parsers
Other API Changes
Deprecations
Removal of prior version deprecations/changes
Performance Improvements
Bug Fixes

1.1.1 New features

Datetime with TZ
We are adding an implementation that natively supports datetime with timezones. A Series or a DataFrame
column previously could be assigned a datetime with timezones, and would work as an object dtype. This had

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

performance issues with a large number rows. See the docs for more details. (GH8260, GH10763, GH11034).
The new implementation allows for having a single-timezone across all rows, with operations in a performant manner.
In [1]: df = DataFrame({'A' : date_range('20130101',periods=3),
...:
'B' : date_range('20130101',periods=3,tz='US/Eastern'),
...:
'C' : date_range('20130101',periods=3,tz='CET')})
...:
In [2]: df
Out[2]:
A
B
C
0 2013-01-01 2013-01-01 00:00:00-05:00 2013-01-01 00:00:00+01:00
1 2013-01-02 2013-01-02 00:00:00-05:00 2013-01-02 00:00:00+01:00
2 2013-01-03 2013-01-03 00:00:00-05:00 2013-01-03 00:00:00+01:00
In [3]: df.dtypes
Out[3]:
A
datetime64[ns]
B
datetime64[ns, US/Eastern]
C
datetime64[ns, CET]
dtype: object
In [4]: df.B
Out[4]:
0
2013-01-01 00:00:00-05:00
1
2013-01-02 00:00:00-05:00
2
2013-01-03 00:00:00-05:00
Name: B, dtype: datetime64[ns, US/Eastern]
In [5]: df.B.dt.tz_localize(None)
Out[5]:
0
2013-01-01
1
2013-01-02
2
2013-01-03
Name: B, dtype: datetime64[ns]

This uses a new-dtype representation as well, that is very similar in look-and-feel to its numpy cousin
datetime64[ns]
In [6]: df['B'].dtype
Out[6]: datetime64[ns, US/Eastern]
In [7]: type(df['B'].dtype)
Out[7]: pandas.core.dtypes.DatetimeTZDtype

Note: There is a slightly different string repr for the underlying DatetimeIndex as a result of the dtype changes,
but functionally these are the same.
Previous Behavior:
In [1]: pd.date_range('20130101',periods=3,tz='US/Eastern')
Out[1]: DatetimeIndex(['2013-01-01 00:00:00-05:00', '2013-01-02 00:00:00-05:00',
'2013-01-03 00:00:00-05:00'],
dtype='datetime64[ns]', freq='D', tz='US/Eastern')
In [2]: pd.date_range('20130101',periods=3,tz='US/Eastern').dtype
Out[2]: dtype('<M8[ns]')

New Behavior:
1.1. v0.17.0 (October 9, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [8]: pd.date_range('20130101',periods=3,tz='US/Eastern')
Out[8]:
DatetimeIndex(['2013-01-01 00:00:00-05:00', '2013-01-02 00:00:00-05:00',
'2013-01-03 00:00:00-05:00'],
dtype='datetime64[ns, US/Eastern]', freq='D')
In [9]: pd.date_range('20130101',periods=3,tz='US/Eastern').dtype
Out[9]: datetime64[ns, US/Eastern]

Releasing the GIL

We are releasing the global-interpreter-lock (GIL) on some cython operations. This will allow other threads to run
simultaneously during computation, potentially allowing performance improvements from multi-threading. Notably
groupby, nsmallest, value_counts and some indexing operations benefit from this. (GH8882)
For example the groupby expression in the following code will have the GIL released during the factorization step,
e.g. df.groupby(key) as well as the .sum() operation.
N = 1000000
ngroups = 10
df = DataFrame({'key' : np.random.randint(0,ngroups,size=N),
'data' : np.random.randn(N) })
df.groupby('key')['data'].sum()

Releasing of the GIL could benefit an application that uses threads for user interactions (e.g. QT), or performing
multi-threaded computations. A nice example of a library that can handle these types of computation-in-parallel is the
dask library.
Plot submethods
The Series and DataFrame .plot() method allows for customizing plot types by supplying the kind keyword
arguments. Unfortunately, many of these kinds of plots use different required and optional keyword arguments, which
makes it difficult to discover what any given plot kind uses out of the dozens of possible arguments.
To alleviate this issue, we have added a new, optional plotting interface, which exposes each kind of plot as a
method of the .plot attribute. Instead of writing series.plot(kind=<kind>, ...), you can now also
use series.plot.<kind>(...):
In [10]: df = pd.DataFrame(np.random.rand(10, 2), columns=['a', 'b'])
In [11]: df.plot.bar()

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

As a result of this change, these methods are now all discoverable via tab-completion:
In [12]: df.plot.<TAB>
df.plot.area
df.plot.barh
df.plot.bar
df.plot.box

df.plot.density
df.plot.hexbin

df.plot.hist
df.plot.kde

df.plot.line
df.plot.pie

df.plot.scatter

Each method signature only includes relevant arguments. Currently, these are limited to required arguments, but in the
future these will include optional arguments, as well. For an overview, see the new Plotting API documentation.
Additional methods for dt accessor
strftime

We are now supporting a Series.dt.strftime method for datetime-likes to generate a formatted string
(GH10110). Examples:
# DatetimeIndex
In [13]: s = pd.Series(pd.date_range('20130101', periods=4))
In [14]: s
Out[14]:
0
2013-01-01
1
2013-01-02
2
2013-01-03
3
2013-01-04
dtype: datetime64[ns]
In [15]: s.dt.strftime('%Y/%m/%d')
Out[15]:
0
2013/01/01
1
2013/01/02
2
2013/01/03
3
2013/01/04
dtype: object

1.1. v0.17.0 (October 9, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

# PeriodIndex
In [16]: s = pd.Series(pd.period_range('20130101', periods=4))
In [17]: s
Out[17]:
0
2013-01-01
1
2013-01-02
2
2013-01-03
3
2013-01-04
dtype: object
In [18]: s.dt.strftime('%Y/%m/%d')
Out[18]:
0
2013/01/01
1
2013/01/02
2
2013/01/03
3
2013/01/04
dtype: object

The string format is as the python standard library and details can be found here
total_seconds

pd.Series of type timedelta64 has new method .dt.total_seconds() returning the duration of the
timedelta in seconds (GH10817)
# TimedeltaIndex
In [19]: s = pd.Series(pd.timedelta_range('1 minutes', periods=4))
In [20]: s
Out[20]:
0
0 days 00:01:00
1
1 days 00:01:00
2
2 days 00:01:00
3
3 days 00:01:00
dtype: timedelta64[ns]
In [21]: s.dt.total_seconds()
Out[21]:
0
60
1
86460
2
172860
3
259260
dtype: float64

Period Frequency Enhancement

Period, PeriodIndex and period_range can now accept multiplied freq. Also, Period.freq and
PeriodIndex.freq are now stored as a DateOffset instance like DatetimeIndex, and not as str
(GH7811)
A multiplied freq represents a span of corresponding length. The example below creates a period of 3 days. Addition
and subtraction will shift the period by its span.
In [22]: p = pd.Period('2015-08-01', freq='3D')

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [23]: p
Out[23]: Period('2015-08-01', '3D')
In [24]: p + 1
Out[24]: Period('2015-08-04', '3D')
In [25]: p - 2
Out[25]: Period('2015-07-26', '3D')
In [26]: p.to_timestamp()
Out[26]: Timestamp('2015-08-01 00:00:00')
In [27]: p.to_timestamp(how='E')
Out[27]: Timestamp('2015-08-03 00:00:00')

You can use the multiplied freq in PeriodIndex and period_range.

In [28]: idx = pd.period_range('2015-08-01', periods=4, freq='2D')

In [29]: idx
Out[29]: PeriodIndex(['2015-08-01', '2015-08-03', '2015-08-05', '2015-08-07'], dtype='int64', freq='2

In [30]: idx + 1
Out[30]: PeriodIndex(['2015-08-03', '2015-08-05', '2015-08-07', '2015-08-09'], dtype='int64', freq='2

Support for SAS XPORT files

read_sas() provides support for reading SAS XPORT format files. (GH4052).
df = pd.read_sas('sas_xport.xpt')

It is also possible to obtain an iterator and read an XPORT file incrementally.

for df in pd.read_sas('sas_xport.xpt', chunksize=10000)
do_something(df)

See the docs for more details.

Support for Math Functions in .eval()
eval() now supports calling math functions (GH4893)
df = pd.DataFrame({'a': np.random.randn(10)})
df.eval("b = sin(a)")

The support math functions are sin, cos, exp, log, expm1, log1p, sqrt, sinh, cosh, tanh, arcsin, arccos, arctan, arccosh,
arcsinh, arctanh, abs and arctan2.
These functions map to the intrinsics for the NumExpr engine. For the Python engine, they are mapped to NumPy
calls.
Changes to Excel with MultiIndex
In version 0.16.2 a DataFrame with MultiIndex columns could not be written to Excel via to_excel. That
functionality has been added (GH10564), along with updating read_excel so that the data can be read back with, no

1.1. v0.17.0 (October 9, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

loss of information, by specifying which columns/rows make up the MultiIndex in the header and index_col
parameters (GH4679)
See the documentation for more details.
In [31]: df = pd.DataFrame([[1,2,3,4], [5,6,7,8]],
....:
columns = pd.MultiIndex.from_product([['foo','bar'],['a','b']],
....:
names = ['col1', 'col2']),
....:
index = pd.MultiIndex.from_product([['j'], ['l', 'k']],
....:
names = ['i1', 'i2']))
....:
In [32]: df
Out[32]:
col1 foo
bar
col2
a b
a
i1 i2
j l
1 2
3
k
5 6
7

b
4
8

In [33]: df.to_excel('test.xlsx')
In [34]: df = pd.read_excel('test.xlsx', header=[0,1], index_col=[0,1])
In [35]: df
Out[35]:
col1 foo
bar
col2
a b
a
i1 i2
j l
1 2
3
k
5 6
7

b
4
8

Previously, it was necessary to specify the has_index_names argument in read_excel, if the serialized data
had index names. For version 0.17.0 the ouptput format of to_excel has been changed to make this keyword
unnecessary - the change is shown below.
Old

New

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Warning: Excel files saved in version 0.16.2 or prior that had index names will still able to be read in, but the
has_index_names argument must specified to True.

Google BigQuery Enhancements

Added ability to automatically create a table/dataset using the pandas.io.gbq.to_gbq() function if the
destination table/dataset does not exist. (GH8325, GH11121).
Added ability to replace an existing table and schema when calling the pandas.io.gbq.to_gbq() function via the if_exists argument. See the docs for more details (GH8325).
InvalidColumnOrder and InvalidPageToken in the gbq module will raise ValueError instead of
IOError.
The generate_bq_schema() function is now deprecated and will be removed in a future version
(GH11121)
The gbq module will now support Python 3 (GH11094).
Display Alignment with Unicode East Asian Width
Warning: Enabling this option will affect the performance for printing of DataFrame and Series (about 2
times slower). Use only when it is actually required.
Some East Asian countries use Unicode characters its width is corresponding to 2 alphabets. If a DataFrame or
Series contains these characters, the default output cannot be aligned properly. The following options are added to
enable precise handling for these characters.
display.unicode.east_asian_width: Whether to use the Unicode East Asian Width to calculate the
display text width. (GH2612)
display.unicode.ambiguous_as_wide: Whether to handle Unicode characters belong to Ambiguous
as Wide. (GH11102)
In [36]: df = pd.DataFrame({u'': ['UK', u''], u'': ['Alice', u'']})
In [37]: df;

1.1. v0.17.0 (October 9, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [38]: pd.set_option('display.unicode.east_asian_width', True)

In [39]: df;

For further details, see here

Other enhancements
Support for openpyxl >= 2.2. The API for style support is now stable (GH10125)
merge now accepts the argument indicator which adds a Categorical-type column (by default called
_merge) to the output object that takes on the values (GH8790)
Observation Origin
Merge key only in left frame
Merge key only in right frame
Merge key in both frames

_merge value
left_only
right_only
both

In [40]: df1 = pd.DataFrame({'col1':[0,1], 'col_left':['a','b']})

In [41]: df2 = pd.DataFrame({'col1':[1,2,2],'col_right':[2,2,2]})
In [42]: pd.merge(df1, df2, on='col1', how='outer', indicator=True)
Out[42]:
col1 col_left col_right
_merge
0
0
a
NaN
left_only
1
1
b
2
both
2
2
NaN
2 right_only
3
2
NaN
2 right_only

For more, see the updated docs

pd.to_numeric is a new function to coerce strings to numbers (possibly with coercion) (GH11133)
pd.merge will now allow duplicate column names if they are not merged upon (GH10639).
pd.pivot will now allow passing index as None (GH3962).
pd.concat will now use existing Series names if provided (GH10698).
In [43]: foo = pd.Series([1,2], name='foo')
In [44]: bar = pd.Series([1,2])
In [45]: baz = pd.Series([4,5])

Previous Behavior:

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [1] pd.concat([foo, bar, baz], 1)

Out[1]:
0 1 2
0 1 1 4
1 2 2 5

New Behavior:
In [46]: pd.concat([foo, bar, baz], 1)
Out[46]:
foo 0 1
0
1 1 4
1
2 2 5

DataFrame has gained the nlargest and nsmallest methods (GH10393)

Add a limit_direction keyword argument that works with limit to enable interpolate to fill NaN
values forward, backward, or both (GH9218, GH10420, GH11115)
In [47]: ser = pd.Series([np.nan, np.nan, 5, np.nan, np.nan, np.nan, 13])
In [48]: ser.interpolate(limit=1, limit_direction='both')
Out[48]:
0
NaN
1
5
2
5
3
7
4
NaN
5
11
6
13
dtype: float64

Added a DataFrame.round method to round the values to a variable number of decimal places (GH10568).
In [49]: df = pd.DataFrame(np.random.random([3, 3]), columns=['A', 'B', 'C'],
....: index=['first', 'second', 'third'])
....:
In [50]: df
Out[50]:
first
second
third

A
0.179356
0.851401
0.597175

B
0.908835
0.203918
0.113613

C
0.571981
0.105336
0.326599

In [51]: df.round(2)
Out[51]:
A
B
C
first
0.18 0.91 0.57
second 0.85 0.20 0.11
third
0.60 0.11 0.33
In [52]: df.round({'A': 0, 'C': 2})
Out[52]:
A
B
C
first
0 0.908835 0.57
second 1 0.203918 0.11
third
1 0.113613 0.33

drop_duplicates and duplicated now accept a keep keyword to target first, last, and all duplicates.
1.1. v0.17.0 (October 9, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

The take_last keyword is deprecated, see here (GH6511, GH8505)

In [53]: s = pd.Series(['A', 'B', 'C', 'A', 'B', 'D'])
In [54]: s.drop_duplicates()
Out[54]:
0
A
1
B
2
C
5
D
dtype: object
In [55]: s.drop_duplicates(keep='last')
Out[55]:
2
C
3
A
4
B
5
D
dtype: object
In [56]: s.drop_duplicates(keep=False)
Out[56]:
2
C
5
D
dtype: object

Reindex now has a tolerance argument that allows for finer control of Limits on filling while reindexing
(GH10411):
In [57]: df = pd.DataFrame({'x': range(5),
....:
't': pd.date_range('2000-01-01', periods=5)})
....:
In [58]: df.reindex([0.1, 1.9, 3.5],
....:
method='nearest',
....:
tolerance=0.2)
....:
Out[58]:
t
x
0.1 2000-01-01
0
1.9 2000-01-03
2
3.5
NaT NaN

When used on a DatetimeIndex, TimedeltaIndex or PeriodIndex, tolerance will coerced into

a Timedelta if possible. This allows you to specify tolerance with a string:
In [59]: df = df.set_index('t')
In [60]: df.reindex(pd.to_datetime(['1999-12-31']),
....:
method='nearest',
....:
tolerance='1 day')
....:
Out[60]:
x
1999-12-31 0

tolerance is also exposed by the lower level Index.get_indexer and Index.get_loc methods.
Added functionality to use the base argument when resampling a TimeDeltaIndex (GH10530)
DatetimeIndex can be instantiated using strings contains NaT (GH7599)
14

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

to_datetime can now accept the yearfirst keyword (GH7599)

pandas.tseries.offsets larger than the Day offset can now be used with a Series for addition/subtraction (GH10699). See the docs for more details.
pd.Timedelta.total_seconds() now returns Timedelta duration to ns precision (previously microsecond precision) (GH10939)
PeriodIndex now supports arithmetic with np.ndarray (GH10638)
Support pickling of Period objects (GH10439)
.as_blocks will now take a copy optional argument to return a copy of the data, default is to copy (no
change in behavior from prior versions), (GH9607)
regex argument to DataFrame.filter now handles numeric column names instead of raising
ValueError (GH10384).
Enable reading gzip compressed files via URL, either by explicitly setting the compression parameter or by
inferring from the presence of the HTTP Content-Encoding header in the response (GH8685)
Enable writing Excel files in memory using StringIO/BytesIO (GH7074)
Enable serialization of lists and dicts to strings in ExcelWriter (GH8188)
SQL io functions now accept a SQLAlchemy connectable. (GH7877)
pd.read_sql and to_sql can accept database URI as con parameter (GH10214)
read_sql_table will now allow reading from views (GH10750).
Enable writing complex values to HDFStores when using the table format (GH10447)
Enable pd.read_hdf to be used without specifying a key when the HDF file contains a single dataset
(GH10443)
pd.read_stata will now read Stata 118 type files. (GH9882)
msgpack submodule has been updated to 0.4.6 with backward compatibility (GH10581)
DataFrame.to_dict now accepts orient=index keyword argument (GH10844).
DataFrame.apply will return a Series of dicts if the passed function returns a dict and reduce=True
(GH8735).
Allow passing kwargs to the interpolation methods (GH10378).
Improved error message when concatenating an empty iterable of Dataframe objects (GH9157)
pd.read_csv can now read bz2-compressed files incrementally, and the C parser can read bz2-compressed
files from AWS S3 (GH11070, GH11072).
In pd.read_csv, recognize s3n:// and s3a:// URLs as designating S3 file storage (GH11070,
GH11071).
Read CSV files from AWS S3 incrementally, instead of first downloading the entire file. (Full file download still
required for compressed files in Python 2.) (GH11070, GH11073)
pd.read_csv is now able to infer compression type for files read from AWS S3 storage (GH11070,
GH11074).

1.1. v0.17.0 (October 9, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.1.2 Backwards incompatible API changes

Changes to sorting API
The sorting API has had some longtime inconsistencies. (GH9816, GH8239).
Here is a summary of the API PRIOR to 0.17.0:
Series.sort is INPLACE while DataFrame.sort returns a new object.
Series.order returns a new object
It was possible to use Series/DataFrame.sort_index to sort by values by passing the by keyword.
Series/DataFrame.sortlevel worked only on a MultiIndex for sorting by index.
To address these issues, we have revamped the API:
We have introduced a new method, DataFrame.sort_values(), which is the merger of
DataFrame.sort(), Series.sort(), and Series.order(), to handle sorting of values.
The existing methods Series.sort(), Series.order(), and DataFrame.sort() have been deprecated and will be removed in a future version.
The by argument of DataFrame.sort_index() has been deprecated and will be removed in a future
version.
The existing method .sort_index() will gain the level keyword to enable level sorting.
We now have two distinct and non-overlapping methods of sorting.
FutureWarning.

A * marks items that will show a

To sort by the values:

Previous
* Series.order()
* Series.sort()
* DataFrame.sort(columns=...)

Replacement
Series.sort_values()
Series.sort_values(inplace=True)
DataFrame.sort_values(by=...)

To sort by the index:

Previous
Series.sort_index()
Series.sortlevel(level=...)
DataFrame.sort_index()
DataFrame.sortlevel(level=...)
* DataFrame.sort()

Replacement
Series.sort_index()
Series.sort_index(level=...)
DataFrame.sort_index()
DataFrame.sort_index(level=...)
DataFrame.sort_index()

We have also deprecated and changed similar methods in two Series-like classes, Index and Categorical.
Previous
* Index.order()
* Categorical.order()

Replacement
Index.sort_values()
Categorical.sort_values()

Changes to to_datetime and to_timedelta

Error handling

The default for pd.to_datetime error handling has changed to errors=raise. In prior versions it was
errors=ignore. Furthermore, the coerce argument has been deprecated in favor of errors=coerce.
This means that invalid parsing will raise rather that return the original input as in previous versions. (GH10636)

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Previous Behavior:
In [2]: pd.to_datetime(['2009-07-31', 'asd'])
Out[2]: array(['2009-07-31', 'asd'], dtype=object)

New Behavior:
In [3]: pd.to_datetime(['2009-07-31', 'asd'])
ValueError: Unknown string format

Of course you can coerce this as well.

In [61]: to_datetime(['2009-07-31', 'asd'], errors='coerce')
Out[61]: DatetimeIndex(['2009-07-31', 'NaT'], dtype='datetime64[ns]', freq=None)

To keep the previous behavior, you can use errors=ignore:

In [62]: to_datetime(['2009-07-31', 'asd'], errors='ignore')
Out[62]: array(['2009-07-31', 'asd'], dtype=object)

Furthermore, pd.to_timedelta has gained a similar API, of errors=raise|ignore|coerce, and

the coerce keyword has been deprecated in favor of errors=coerce.
Consistent Parsing

The string parsing of to_datetime, Timestamp and DatetimeIndex has been made consistent. (GH7599)
Prior to v0.17.0, Timestamp and to_datetime may parse year-only datetime-string incorrectly using todays
date, otherwise DatetimeIndex uses the beginning of the year. Timestamp and to_datetime may raise
ValueError in some types of datetime-string which DatetimeIndex can parse, such as a quarterly string.
Previous Behavior:
In [1]: Timestamp('2012Q2')
Traceback
...
ValueError: Unable to parse 2012Q2
# Results in today's date.
In [2]: Timestamp('2014')
Out [2]: 2014-08-12 00:00:00

v0.17.0 can parse them as below. It works on DatetimeIndex also.

New Behavior:
In [63]: Timestamp('2012Q2')
Out[63]: Timestamp('2012-04-01 00:00:00')
In [64]: Timestamp('2014')
Out[64]: Timestamp('2014-01-01 00:00:00')
In [65]: DatetimeIndex(['2012Q2', '2014'])
Out[65]: DatetimeIndex(['2012-04-01', '2014-01-01'], dtype='datetime64[ns]', freq=None)

Note:
If you want to perform calculations based on todays date, use Timestamp.now() and
pandas.tseries.offsets.

1.1. v0.17.0 (October 9, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [66]: import pandas.tseries.offsets as offsets

In [67]: Timestamp.now()
Out[67]: Timestamp('2015-10-09 20:59:45.917782')
In [68]: Timestamp.now() + offsets.DateOffset(years=1)
Out[68]: Timestamp('2016-10-09 20:59:45.919984')

Changes to Index Comparisons

Operator equal on Index should behavior similarly to Series (GH9947, GH10637)
Starting in v0.17.0, comparing Index objects of different lengths will raise a ValueError. This is to be consistent
with the behavior of Series.
Previous Behavior:
In [2]: pd.Index([1, 2, 3]) == pd.Index([1, 4, 5])
Out[2]: array([ True, False, False], dtype=bool)
In [3]: pd.Index([1, 2, 3]) == pd.Index([2])
Out[3]: array([False, True, False], dtype=bool)
In [4]: pd.Index([1, 2, 3]) == pd.Index([1, 2])
Out[4]: False

New Behavior:
In [8]: pd.Index([1, 2, 3]) == pd.Index([1, 4, 5])
Out[8]: array([ True, False, False], dtype=bool)
In [9]: pd.Index([1, 2, 3]) == pd.Index([2])
ValueError: Lengths must match to compare
In [10]: pd.Index([1, 2, 3]) == pd.Index([1, 2])
ValueError: Lengths must match to compare

Note that this is different from the numpy behavior where a comparison can be broadcast:
In [69]: np.array([1, 2, 3]) == np.array([1])
Out[69]: array([ True, False, False], dtype=bool)

or it can return False if broadcasting can not be done:

In [70]: np.array([1, 2, 3]) == np.array([1, 2])
Out[70]: False

Changes to Boolean Comparisons vs. None

Boolean comparisons of a Series vs None will now be equivalent to comparing with np.nan, rather than raise
TypeError. (GH1079).
In [71]: s = Series(range(3))
In [72]: s.iloc[1] = None

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [73]: s
Out[73]:
0
0
1
NaN
2
2
dtype: float64

Previous Behavior:
In [5]: s==None
TypeError: Could not compare <type 'NoneType'> type with Series

New Behavior:
In [74]: s==None
Out[74]:
0
False
1
False
2
False
dtype: bool

Usually you simply want to know which values are null.

In [75]: s.isnull()
Out[75]:
0
False
1
True
2
False
dtype: bool

Warning:
You generally will want to use isnull/notnull for these types of comparisons, as
isnull/notnull tells you which elements are null. One has to be mindful that nans dont compare equal,
but Nones do. Note that Pandas/numpy uses the fact that np.nan != np.nan, and treats None like np.nan.
In [76]: None == None
Out[76]: True
In [77]: np.nan == np.nan
Out[77]: False

HDFStore dropna behavior

The default behavior for HDFStore write functions with format=table is now to keep rows that are all missing.
Previously, the behavior was to drop rows that were all missing save the index. The previous behavior can be replicated
using the dropna=True option. (GH9382)
Previous Behavior:
In [78]: df_with_missing = pd.DataFrame({'col1':[0, np.nan, 2],
....:
'col2':[1, np.nan, np.nan]})
....:
In [79]: df_with_missing
Out[79]:
col1 col2
0
0
1

1.1. v0.17.0 (October 9, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

1
2

NaN
2

NaN
NaN

In [28]:
df_with_missing.to_hdf('file.h5',
'df_with_missing',
format='table',
mode='w')
pd.read_hdf('file.h5', 'df_with_missing')
Out [28]:
col1
0
0
2
2

col2
1
NaN

New Behavior:
In [80]: df_with_missing.to_hdf('file.h5',
....:
'df_with_missing',
....:
format='table',
....:
mode='w')
....:
In [81]: pd.read_hdf('file.h5', 'df_with_missing')
Out[81]:
col1 col2
0
0
1
1
NaN
NaN
2
2
NaN

See the docs for more details.

Changes to display.precision option
The display.precision option has been clarified to refer to decimal places (GH10451).
Earlier versions of pandas would format floating point numbers to have one less decimal place than the value in
display.precision.
In [1]: pd.set_option('display.precision', 2)
In [2]: pd.DataFrame({'x': [123.456789]})
Out[2]:
x
0 123.5

If interpreting precision as significant figures this did work for scientific notation but that same interpretation did not
work for values with standard formatting. It was also out of step with how numpy handles formatting.
Going forward the value of display.precision will directly control the number of places after the decimal, for
regular formatting as well as scientific notation, similar to how numpys precision print option works.
In [82]: pd.set_option('display.precision', 2)
In [83]: pd.DataFrame({'x': [123.456789]})
Out[83]:
x
0 123.46

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

To preserve output behavior with prior versions the default value of display.precision has been reduced to 6
from 7.
Changes to Categorical.unique
Categorical.unique now returns new Categoricals with categories and codes that are unique, rather
than returning np.array (GH10508)
unordered category: values and categories are sorted by appearance order.
ordered category: values are sorted by appearance order, categories keep existing order.
In [84]: cat = pd.Categorical(['C', 'A', 'B', 'C'],
....:
categories=['A', 'B', 'C'],
....:
ordered=True)
....:
In [85]: cat
Out[85]:
[C, A, B, C]
Categories (3, object): [A < B < C]
In [86]: cat.unique()
Out[86]:
[C, A, B]
Categories (3, object): [A < B < C]
In [87]: cat = pd.Categorical(['C', 'A', 'B', 'C'],
....:
categories=['A', 'B', 'C'])
....:
In [88]: cat
Out[88]:
[C, A, B, C]
Categories (3, object): [A, B, C]
In [89]: cat.unique()
Out[89]:
[C, A, B]
Categories (3, object): [C, A, B]

Changes to bool passed as header in Parsers

In earlier versions of pandas, if a bool was passed the header argument of read_csv, read_excel, or
read_html it was implicitly converted to an integer, resulting in header=0 for False and header=1 for True
(GH6113)
A bool input to header will now raise a TypeError
In [29]: df = pd.read_csv('data.csv', header=False)
TypeError: Passing a bool to header is invalid. Use header=None for no header or
header=int or list-like of ints to specify the row(s) making up the column names

1.1. v0.17.0 (October 9, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Other API Changes

Line and kde plot with subplots=True now uses default colors, not all black. Specify color=k to draw
all lines in black (GH9894)
Calling the .value_counts() method on a Series with a categorical dtype now returns a Series with a
CategoricalIndex (GH10704)
The metadata properties of subclasses of pandas objects will now be serialized (GH10553).
groupby using Categorical follows the same rule as Categorical.unique described above
(GH10508)
When constructing DataFrame with an array of complex64 dtype previously meant the corresponding column was automatically promoted to the complex128 dtype. Pandas will now preserve the itemsize of the
input for complex data (GH10952)
some numeric reduction operators would return ValueError, rather than TypeError on object types that
includes strings and numbers (GH11131)
Passing currently unsupported chunksize argument to read_excel or ExcelFile.parse will now
raise NotImplementedError (GH8011)
Allow an ExcelFile object to be passed into read_excel (GH11198)
DatetimeIndex.union does not infer freq if self and the input have None as freq (GH11086)
NaTs methods now either raise ValueError, or return np.nan or NaT (GH9513)
Behavior
return np.nan
return NaT
return np.datetime64(NaT)
raise ValueError

Methods
weekday, isoweekday
date, now, replace, to_datetime, today
to_datetime64 (unchanged)
All other public methods (names not beginning with underscores)

Deprecations
For Series the following indexing functions are deprecated (GH10177).
Deprecated Function
.irow(i)
.iget(i)
.iget_value(i)

Replacement
.iloc[i] or .iat[i]
.iloc[i] or .iat[i]
.iloc[i] or .iat[i]

For DataFrame the following indexing functions are deprecated (GH10177).

Deprecated Function
.irow(i)
.iget_value(i, j)
.icol(j)

Replacement
.iloc[i]
.iloc[i, j] or .iat[i, j]
.iloc[:, j]

Note: These indexing function have been deprecated in the documentation since 0.11.0.
Categorical.name was deprecated to make Categorical more numpy.ndarray like.
Series(cat, name="whatever") instead (GH10482).

Use

Setting missing values (NaN) in a Categoricals categories will issue a warning (GH10748). You can
still have missing values in the values.
drop_duplicates and duplicateds take_last keyword was deprecated in favor of keep. (GH6511,
GH8505)

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Series.nsmallest and nlargests take_last keyword was deprecated in favor of keep. (GH10792)
DataFrame.combineAdd and DataFrame.combineMult are deprecated. They can easily be
replaced by using the add and mul methods: DataFrame.add(other, fill_value=0) and
DataFrame.mul(other, fill_value=1.) (GH10735).
TimeSeries deprecated in favor of Series (note that this has been an alias since 0.13.0), (GH10890)
SparsePanel deprecated and will be removed in a future version (GH11157).
Series.is_time_series deprecated in favor of Series.index.is_all_dates (GH11135)
Legacy offsets (like A@JAN) listed in here are deprecated (note that this has been alias since 0.8.0),
(GH10878)
WidePanel deprecated in favor of Panel, LongPanel in favor of DataFrame (note these have been
aliases since < 0.11.0), (GH10892)
DataFrame.convert_objects has been deprecated in favor of type-specific
pd.to_datetime, pd.to_timestamp and pd.to_numeric (new in 0.17.0) (GH11133).

functions

Removal of prior version deprecations/changes

Removal of na_last parameters from Series.order() and Series.sort(), in favor of
na_position. (GH5231)
Remove of percentile_width from .describe(), in favor of percentiles. (GH7088)
Removal of colSpace parameter from DataFrame.to_string(), in favor of col_space, circa 0.8.0
version.
Removal of automatic time-series broadcasting (GH2304)
In [90]: np.random.seed(1234)

In [91]: df = DataFrame(np.random.randn(5,2),columns=list('AB'),index=date_range('20130101',peri
In [92]: df
Out[92]:
A
B
2013-01-01 0.471435 -1.190976
2013-01-02 1.432707 -0.312652
2013-01-03 -0.720589 0.887163
2013-01-04 0.859588 -0.636524
2013-01-05 0.015696 -2.242685

Previously
In [3]: df + df.A
FutureWarning: TimeSeries broadcasting along DataFrame index by default is deprecated.
Please use DataFrame.<op> to explicitly broadcast arithmetic operations along the index
Out[3]:
A
B
2013-01-01 0.942870 -0.719541
2013-01-02 2.865414 1.120055
2013-01-03 -1.441177 0.166574
2013-01-04 1.719177 0.223065
2013-01-05 0.031393 -2.226989

Current

1.1. v0.17.0 (October 9, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [93]: df.add(df.A,axis='index')
Out[93]:
A
B
2013-01-01 0.942870 -0.719541
2013-01-02 2.865414 1.120055
2013-01-03 -1.441177 0.166574
2013-01-04 1.719177 0.223065
2013-01-05 0.031393 -2.226989

Remove table keyword in HDFStore.put/append, in favor of using format= (GH4645)

Remove kind in read_excel/ExcelFile as its unused (GH4712)
Remove infer_type keyword from pd.read_html as its unused (GH4770, GH7032)
Remove offset and timeRule keywords from Series.tshift/shift, in favor of freq (GH4853,
GH4864)
Remove pd.load/pd.save aliases in favor of pd.to_pickle/pd.read_pickle (GH3787)

1.1.3 Performance Improvements

Development support for benchmarking with the Air Speed Velocity library (GH8361)
Added vbench benchmarks for alternative ExcelWriter engines and reading Excel files (GH7171)
Performance improvements in Categorical.value_counts (GH10804)
Performance improvements in SeriesGroupBy.nunique and SeriesGroupBy.value_counts and
SeriesGroupby.transform (GH10820, GH11077)
Performance improvements in DataFrame.drop_duplicates with integer dtypes (GH10917)
Performance improvements in DataFrame.duplicated with wide frames. (GH10161, GH11180)
4x improvement in timedelta string parsing (GH6755, GH10426)
8x improvement in timedelta64 and datetime64 ops (GH6755)
Significantly improved performance of indexing MultiIndex with slicers (GH10287)
8x improvement in iloc using list-like input (GH10791)
Improved performance of Series.isin for datetimelike/integer Series (GH10287)
20x improvement in concat of Categoricals when categories are identical (GH10587)
Improved performance of to_datetime when specified format string is ISO8601 (GH10178)
2x improvement of Series.value_counts for float dtype (GH10821)
Enable infer_datetime_format in to_datetime when date components do not have 0 padding
(GH11142)
Regression from 0.16.1 in constructing DataFrame from nested dictionary (GH11084)
Performance improvements in addition/subtraction operations for DateOffset with Series or
DatetimeIndex (GH10744, GH11205)

1.1.4 Bug Fixes

Bug in incorrection computation of .mean() on timedelta64[ns] because of overflow (GH9442)

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug in .isin on older numpies (:issue: 11232)

Bug in DataFrame.to_html(index=False) renders unnecessary name row (GH10344)
Bug in DataFrame.to_latex() the column_format argument could not be passed (GH9402)
Bug in DatetimeIndex when localizing with NaT (GH10477)
Bug in Series.dt ops in preserving meta-data (GH10477)
Bug in preserving NaT when passed in an otherwise invalid to_datetime construction (GH10477)
Bug in DataFrame.apply when function returns categorical series. (GH9573)
Bug in to_datetime with invalid dates and formats supplied (GH10154)
Bug in Index.drop_duplicates dropping name(s) (GH10115)
Bug in Series.quantile dropping name (GH10881)
Bug in pd.Series when setting a value on an empty Series whose index has a frequency. (GH10193)
Bug in pd.Series.interpolate with invalid order keyword values. (GH10633)
Bug in DataFrame.plot raises ValueError when color name is specified by multiple characters
(GH10387)
Bug in Index construction with a mixed list of tuples (GH10697)
Bug in DataFrame.reset_index when index contains NaT. (GH10388)
Bug in ExcelReader when worksheet is empty (GH6403)
Bug in BinGrouper.group_info where returned values are not compatible with base class (GH10914)
Bug in clearing the cache on DataFrame.pop and a subsequent inplace op (GH10912)
Bug in indexing with a mixed-integer Index causing an ImportError (GH10610)
Bug in Series.count when index has nulls (GH10946)
Bug in pickling of a non-regular freq DatetimeIndex (GH11002)
Bug causing DataFrame.where to not respect the axis parameter when the frame has a symmetric shape.
(GH9736)
Bug in Table.select_column where name is not preserved (GH10392)
Bug in offsets.generate_range where start and end have finer precision than offset (GH9907)
Bug in pd.rolling_* where Series.name would be lost in the output (GH10565)
Bug in stack when index or columns are not unique. (GH10417)
Bug in setting a Panel when an axis has a multi-index (GH10360)
Bug in USFederalHolidayCalendar where USMemorialDay and USMartinLutherKingJr were
incorrect (GH10278 and GH9760 )
Bug in .sample() where returned object, if set, gives unnecessary SettingWithCopyWarning
(GH10738)
Bug in .sample() where weights passed as Series were not aligned along axis before being treated positionally, potentially causing problems if weight indices were not aligned with sampled object. (GH10738)
Regression fixed in (GH9311, GH6620, GH9345), where groupby with a datetime-like converting to float with
certain aggregators (GH10979)
Bug in DataFrame.interpolate with axis=1 and inplace=True (GH10395)

1.1. v0.17.0 (October 9, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug in io.sql.get_schema when specifying multiple columns as primary key (GH10385).

Bug in groupby(sort=False) with datetime-like Categorical raises ValueError (GH10505)
Bug in groupby(axis=1) with filter() throws IndexError (GH11041)
Bug in test_categorical on big-endian builds (GH10425)
Bug in Series.shift and DataFrame.shift not supporting categorical data (GH9416)
Bug in Series.map using categorical Series raises AttributeError (GH10324)
Bug in MultiIndex.get_level_values including Categorical raises AttributeError
(GH10460)
Bug in pd.get_dummies with sparse=True not returning SparseDataFrame (GH10531)
Bug in Index subtypes (such as PeriodIndex) not returning their own type for .drop and .insert
methods (GH10620)
Bug in algos.outer_join_indexer when right array is empty (GH10618)
Bug in filter (regression from 0.16.0) and transform when grouping on multiple keys, one of which is
datetime-like (GH10114)
Bug in to_datetime and to_timedelta causing Index name to be lost (GH10875)
Bug in len(DataFrame.groupby) causing IndexError when theres a column containing only NaNs
(:issue: 11016)
Bug that caused segfault when resampling an empty Series (GH10228)
Bug in DatetimeIndex and PeriodIndex.value_counts resets name from its result, but retains in
results Index. (GH10150)
Bug in pd.eval using numexpr engine coerces 1 element numpy array to scalar (GH10546)
Bug in pd.concat with axis=0 when column is of dtype category (GH10177)
Bug in read_msgpack where input type is not always checked (GH10369, GH10630)
Bug in pd.read_csv with kwargs index_col=False, index_col=[a, b] or dtype
(GH10413, GH10467, GH10577)
Bug in Series.from_csv with
Series.index.name (GH10483)

header

kwarg

not

setting

the

Series.name

the

Bug in groupby.var which caused variance to be inaccurate for small float values (GH10448)
Bug in Series.plot(kind=hist) Y Label not informative (GH10485)
Bug in read_csv when using a converter which generates a uint8 type (GH9266)
Bug causes memory leak in time-series line and area plot (GH9003)
Bug when setting a Panel sliced along the major or minor axes when the right-hand side is a DataFrame
(GH11014)
Bug that returns None and does not raise NotImplementedError when operator functions (e.g. .add) of
Panel are not implemented (GH7692)
Bug in line and kde plot cannot accept multiple colors when subplots=True (GH9894)
Bug in DataFrame.plot raises ValueError when color name is specified by multiple characters
(GH10387)
Bug in left and right align of Series with MultiIndex may be inverted (GH10665)

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug in left and right join of with MultiIndex may be inverted (GH10741)
Bug in read_stata when reading a file with a different order set in columns (GH10757)
Bug in Categorical may not representing properly when category contains tz or Period (GH10713)
Bug in Categorical.__iter__ may not returning correct datetime and Period (GH10713)
Bug in indexing with a PeriodIndex on an object with a PeriodIndex (GH4125)
Bug in read_csv with engine=c: EOF preceded by a comment, blank line, etc. was not handled correctly
(GH10728, GH10548)
Reading famafrench data via DataReader results in HTTP 404 error because of the website url is changed
(GH10591).
Bug in read_msgpack where DataFrame to decode has duplicate column names (GH9618)
Bug in io.common.get_filepath_or_buffer which caused reading of valid S3 files to fail if the
bucket also contained keys for which the user does not have read permission (GH10604)
Bug in vectorised setting of timestamp columns with python datetime.date and numpy datetime64
(GH10408, GH10412)
Bug in Index.take may add unnecessary freq attribute (GH10791)
Bug in merge with empty DataFrame may raise IndexError (GH10824)
Bug in to_latex where unexpected keyword argument for some documented arguments (GH10888)
Bug in indexing of large DataFrame where IndexError is uncaught (GH10645 and GH10692)
Bug in read_csv when using the nrows or chunksize parameters if file contains only a header line
(GH9535)
Bug in serialization of category types in HDF5 in presence of alternate encodings. (GH10366)
Bug in pd.DataFrame when constructing an empty DataFrame with a string dtype (GH9428)
Bug in pd.DataFrame.diff when DataFrame is not consolidated (GH10907)
Bug in pd.unique for arrays with the datetime64 or timedelta64 dtype that meant an array with object
dtype was returned instead the original dtype (GH9431)
Bug in Timedelta raising error when slicing from 0s (GH10583)
Bug in DatetimeIndex.take and TimedeltaIndex.take may not raise IndexError against invalid
index (GH10295)
Bug in Series([np.nan]).astype(M8[ms]),
(GH10747)

which now returns Series([pd.NaT])

Bug in PeriodIndex.order reset freq (GH10295)

Bug in date_range when freq divides end as nanos (GH10885)
Bug in iloc allowing memory outside bounds of a Series to be accessed with negative integers (GH10779)
Bug in read_msgpack where encoding is not respected (GH10581)
Bug preventing access to the first index when using iloc with a list containing the appropriate negative integer
(GH10547, GH10779)
Bug in TimedeltaIndex formatter causing error while trying to save DataFrame with
TimedeltaIndex using to_csv (GH10833)
Bug in DataFrame.where when handling Series slicing (GH10218, GH9558)

1.1. v0.17.0 (October 9, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug where pd.read_gbq throws ValueError when Bigquery returns zero rows (GH10273)
Bug in to_json which was causing segmentation fault when serializing 0-rank ndarray (GH9576)
Bug in plotting functions may raise IndexError when plotted on GridSpec (GH10819)
Bug in plot result may show unnecessary minor ticklabels (GH10657)
Bug in groupby incorrect computation for aggregation on DataFrame with NaT (E.g first, last, min).
(GH10590, GH11010)
Bug when constructing DataFrame where passing a dictionary with only scalar values and specifying columns
did not raise an error (GH10856)
Bug in .var() causing roundoff errors for highly similar values (GH10242)
Bug in DataFrame.plot(subplots=True) with duplicated columns outputs incorrect result (GH10962)
Bug in Index arithmetic may result in incorrect class (GH10638)
Bug in date_range results in empty if freq is negative annualy, quarterly and monthly (GH11018)
Bug in DatetimeIndex cannot infer negative freq (GH11018)
Remove use of some deprecated numpy comparison operations, mainly in tests. (GH10569)
Bug in Index dtype may not applied properly (GH11017)
Bug in io.gbq when testing for minimum google api client version (GH10652)
Bug in DataFrame construction from nested dict with timedelta keys (GH11129)
Bug in .fillna against may raise TypeError when data contains datetime dtype (GH7095, GH11153)
Bug in .groupby when number of keys to group by is same as length of index (GH11185)
Bug in convert_objects where converted values might not be returned if all null and coerce (GH9589)
Bug in convert_objects where copy keyword was not respected (GH9589)

1.2 v0.16.2 (June 12, 2015)

This is a minor bug-fix release from 0.16.1 and includes a a large number of bug fixes along some new features
(pipe() method), enhancements, and performance improvements.
We recommend that all users upgrade to this version.
Highlights include:
A new pipe method, see here
Documentation on how to use numba with pandas, see here
Whats new in v0.16.2
New features
Pipe
Other Enhancements
API Changes
Performance Improvements
Bug Fixes

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.2.1 New features

Pipe
Weve introduced a new method DataFrame.pipe(). As suggested by the name, pipe should be used to pipe
data through a chain of function calls. The goal is to avoid confusing nested function calls like
# df is a DataFrame
# f, g, and h are functions that take and return DataFrames
f(g(h(df), arg1=1), arg2=2, arg3=3)

The logic flows from inside out, and function names are separated from their keyword arguments. This can be rewritten
as
(df.pipe(h)
.pipe(g, arg1=1)
.pipe(f, arg2=2, arg3=3)
)

Now both the code and the logic flow from top to bottom. Keyword arguments are next to their functions. Overall the
code is much more readable.
In the example above, the functions f, g, and h each expected the DataFrame as the first positional argument. When
the function you wish to apply takes its data anywhere other than the first argument, pass a tuple of (function,
keyword) indicating where the DataFrame should flow. For example:
In [1]: import statsmodels.formula.api as sm
In [2]: bb = pd.read_csv('data/baseball.csv', index_col='id')
# sm.poisson takes (formula, data)
In [3]: (bb.query('h > 0')
...:
.assign(ln_h = lambda df: np.log(df.h))
...:
.pipe((sm.poisson, 'data'), 'hr ~ ln_h + year + g + C(lg)')
...:
.fit()
...:
.summary()
...: )
...:
Optimization terminated successfully.
Current function value: 2.116284
Iterations 24
Out[3]:
<class 'statsmodels.iolib.summary.Summary'>
"""
Poisson Regression Results
==============================================================================
Dep. Variable:
hr
No. Observations:
68
Model:
Poisson
Df Residuals:
63
Method:
MLE
Df Model:
4
Date:
Fri, 09 Oct 2015
Pseudo R-squ.:
0.6878
Time:
20:59:49
Log-Likelihood:
-143.91
converged:
True
LL-Null:
-460.91
LLR p-value:
6.774e-136
===============================================================================
coef
std err
z
P>|z|
[95.0% Conf. Int.]
------------------------------------------------------------------------------Intercept
-1267.3636
457.867
-2.768
0.006
-2164.767 -369.960
C(lg)[T.NL]
-0.2057
0.101
-2.044
0.041
-0.403
-0.008
ln_h
0.9280
0.191
4.866
0.000
0.554
1.302

1.2. v0.16.2 (June 12, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

year
0.6301
0.228
2.762
0.006
0.183
1.077
g
0.0099
0.004
2.754
0.006
0.003
0.017
===============================================================================
"""

The pipe method is inspired by unix pipes, which stream text through processes. More recently dplyr and magrittr
have introduced the popular (%>%) pipe operator for R.
See the documentation for more. (GH10129)
Other Enhancements
Added rsplit to Index/Series StringMethods (GH10303)
Removed the hard-coded size limits on the DataFrame HTML representation in the IPython notebook, and
leave this to IPython itself (only for IPython v3.0 or greater). This eliminates the duplicate scroll bars that
appeared in the notebook with large frames (GH10231).
Note that the notebook has a toggle output scrolling feature to limit the display of very large frames
(by clicking left of the output). You can also configure the way DataFrames are displayed using the pandas
options, see here here.
axis parameter of DataFrame.quantile now accepts also index and column. (GH9543)

1.2.2 API Changes

Holiday now raises NotImplementedError if both offset and observance are used in the constructor instead of returning an incorrect result (GH10217).

1.2.3 Performance Improvements

Improved Series.resample performance with dtype=datetime64[ns] (GH7754)
Increase performance of str.split when expand=True (GH10081)

1.2.4 Bug Fixes

Bug in Series.hist raises an error when a one row Series was given (GH10214)
Bug where HDFStore.select modifies the passed columns list (GH7212)
Bug in Categorical repr with display.width of None in Python 3 (GH10087)
Bug in to_json with certain orients and a CategoricalIndex would segfault (GH10317)
Bug where some of the nan funcs do not have consistent return dtypes (GH10251)
Bug in DataFrame.quantile on checking that a valid axis was passed (GH9543)
Bug in groupby.apply aggregation for Categorical not preserving categories (GH10138)
Bug in to_csv where date_format is ignored if the datetime is fractional (GH10209)
Bug in DataFrame.to_json with mixed data types (GH10289)
Bug in cache updating when consolidating (GH10264)
Bug in mean() where integer dtypes can overflow (GH10172)

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug where Panel.from_dict does not set dtype when specified (GH10058)
Bug in Index.union raises AttributeError when passing array-likes. (GH10149)
Bug in Timestamps microsecond, quarter, dayofyear, week and daysinmonth properties return np.int type, not built-in int. (GH10050)
Bug in NaT raises AttributeError when accessing to daysinmonth, dayofweek properties.
(GH10096)
Bug in Index repr when using the max_seq_items=None setting (GH10182).
Bug in getting timezone data with dateutil on various platforms ( GH9059, GH8639, GH9663, GH10121)
Bug in displaying datetimes with mixed frequencies; display ms datetimes to the proper precision. (GH10170)
Bug in setitem where type promotion is applied to the entire block (GH10280)
Bug in Series arithmetic methods may incorrectly hold names (GH10068)
Bug in GroupBy.get_group when grouping on multiple keys, one of which is categorical. (GH10132)
Bug in DatetimeIndex and TimedeltaIndex names are lost after timedelta arithmetics ( GH9926)
Bug in DataFrame construction from nested dict with datetime64 (GH10160)
Bug in Series construction from dict with datetime64 keys (GH9456)
Bug in Series.plot(label="LABEL") not correctly setting the label (GH10119)
Bug in plot not defaulting to matplotlib axes.grid setting (GH9792)
Bug causing strings containing an exponent, but no decimal to be parsed as int instead of float in
engine=python for the read_csv parser (GH9565)
Bug in Series.align resets name when fill_value is specified (GH10067)
Bug in read_csv causing index name not to be set on an empty DataFrame (GH10184)
Bug in SparseSeries.abs resets name (GH10241)
Bug in TimedeltaIndex slicing may reset freq (GH10292)
Bug in GroupBy.get_group raises ValueError when group key contains NaT (GH6992)
Bug in SparseSeries constructor ignores input data name (GH10258)
Bug in Categorical.remove_categories causing a ValueError when removing the NaN category
if underlying dtype is floating-point (GH10156)
Bug where infer_freq infers timerule (WOM-5XXX) unsupported by to_offset (GH9425)
Bug in DataFrame.to_hdf() where table format would raise a seemingly unrelated error for invalid (nonstring) column names. This is now explicitly forbidden. (GH9057)
Bug to handle masking empty DataFrame (GH10126).
Bug where MySQL interface could not handle numeric table/column names (GH10255)
Bug in read_csv with a date_parser that returned a datetime64 array of other time resolution than
[ns] (GH10245)
Bug in Panel.apply when the result has ndim=0 (GH10332)
Bug in read_hdf where auto_close could not be passed (GH9327).
Bug in read_hdf where open stores could not be used (GH10330).

1.2. v0.16.2 (June 12, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug in adding empty DataFrames, now results in a DataFrame that .equals an empty
DataFrame (GH10181).
Bug in to_hdf and HDFStore which did not check that complib choices were valid (GH4582, GH8874).

1.3 v0.16.1 (May 11, 2015)

This is a minor bug-fix release from 0.16.0 and includes a a large number of bug fixes along several new features,
enhancements, and performance improvements. We recommend that all users upgrade to this version.
Highlights include:
Support for a CategoricalIndex, a category based index, see here
New section on how-to-contribute to pandas, see here
Revised Merge, join, and concatenate documentation, including graphical examples to make it easier to understand each operations, see here
New method sample for drawing random samples from Series, DataFrames and Panels. See here
The default Index printing has changed to a more uniform format, see here
BusinessHour datetime-offset is now supported, see here
Further enhancement to the .str accessor to make string operations easier, see here
Whats new in v0.16.1
Enhancements
CategoricalIndex
Sample
String Methods Enhancements
Other Enhancements
API changes
Deprecations
Index Representation
Performance Improvements
Bug Fixes
Warning: In pandas 0.17.0, the sub-package pandas.io.data will be removed in favor of a separately
installable package. See here for details (GH8961)

1.3.1 Enhancements
CategoricalIndex
We introduce a CategoricalIndex, a new type of index object that is useful for supporting indexing with duplicates. This is a container around a Categorical (introduced in v0.15.0) and allows efficient indexing and storage
of an index with a large number of duplicated elements. Prior to 0.16.1, setting the index of a DataFrame/Series
with a category dtype would convert this to regular object-based Index.

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [1]: df = DataFrame({'A' : np.arange(6),

...:
'B' : Series(list('aabbca')).astype('category',
...:
categories=list('cab'))
...:
})
...:
In [2]: df
Out[2]:
A B
0 0 a
1 1 a
2 2 b
3 3 b
4 4 c
5 5 a
In [3]: df.dtypes
Out[3]:
A
int32
B
category
dtype: object
In [4]: df.B.cat.categories
Out[4]: Index([u'c', u'a', u'b'], dtype='object')

setting the index, will create create a CategoricalIndex

In [5]: df2 = df.set_index('B')

In [6]: df2.index
Out[6]: CategoricalIndex([u'a', u'a', u'b', u'b', u'c', u'a'], categories=[u'c', u'a', u'b'], ordered

indexing with __getitem__/.iloc/.loc/.ix works similarly to an Index with duplicates. The indexers
MUST be in the category or the operation will raise.
In [7]: df2.loc['a']
Out[7]:
A
B
a 0
a 1
a 5

and preserves the CategoricalIndex

In [8]: df2.loc['a'].index
Out[8]: CategoricalIndex([u'a', u'a', u'a'], categories=[u'c', u'a', u'b'], ordered=False, name=u'B',

1.3. v0.16.1 (May 11, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

groupby operations on the index will preserve the index nature as well
In [10]: df2.groupby(level=0).sum()
Out[10]:
A
B
c 4
a 6
b 5

In [11]: df2.groupby(level=0).sum().index
Out[11]: CategoricalIndex([u'c', u'a', u'b'], categories=[u'c', u'a', u'b'], ordered=False, name=u'B'

reindexing operations, will return a resulting index based on the type of the passed indexer, meaning that passing a
list will return a plain-old-Index; indexing with a Categorical will return a CategoricalIndex, indexed
according to the categories of the PASSED Categorical dtype. This allows one to arbitrarly index these even with
values NOT in the categories, similarly to how you can reindex ANY pandas index.
In [12]: df2.reindex(['a','e'])
Out[12]:
A
B
a
0
a
1
a
5
e NaN
In [13]: df2.reindex(['a','e']).index
Out[13]: Index([u'a', u'a', u'a', u'e'], dtype='object', name=u'B')
In [14]: df2.reindex(pd.Categorical(['a','e'],categories=list('abcde')))
Out[14]:
A
B
a
0
a
1
a
5
e NaN

In [15]: df2.reindex(pd.Categorical(['a','e'],categories=list('abcde'))).index
Out[15]: CategoricalIndex([u'a', u'a', u'a', u'e'], categories=[u'a', u'e'], ordered=False, name=u'B'

See the documentation for more. (GH7629, GH10038, GH10039)

Sample
Series, DataFrames, and Panels now have a new method: sample(). The method accepts a specific number of rows
or columns to return, or a fraction of the total number or rows or columns. It also has options for sampling with or
without replacement, for passing in a column for weights for non-uniform sampling, and for setting seed values to
facilitate replication. (GH2419)
In [16]: example_series = Series([0,1,2,3,4,5])
# When no arguments are passed, returns 1
In [17]: example_series.sample()
Out[17]:
0
0
dtype: int64

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

# One may specify either a number of rows:

In [18]: example_series.sample(n=3)
Out[18]:
4
4
2
2
0
0
dtype: int64
# Or a fraction of the rows:
In [19]: example_series.sample(frac=0.5)
Out[19]:
4
4
5
5
3
3
dtype: int64
# weights are accepted.
In [20]: example_weights = [0, 0, 0.2, 0.2, 0.2, 0.4]
In [21]: example_series.sample(n=3, weights=example_weights)
Out[21]:
4
4
2
2
5
5
dtype: int64
# weights will also be normalized if they do not sum to one,
# and missing values will be treated as zeros.
In [22]: example_weights2 = [0.5, 0, 0, 0, None, np.nan]
In [23]: example_series.sample(n=1, weights=example_weights2)
Out[23]:
0
0
dtype: int64

When applied to a DataFrame, one may pass the name of a column to specify sampling weights when sampling from
rows.
In [24]: df = DataFrame({'col1':[9,8,7,6], 'weight_column':[0.5, 0.4, 0.1, 0]})
In [25]: df.sample(n=3, weights='weight_column')
Out[25]:
col1 weight_column
1
8
0.4
0
9
0.5
2
7
0.1

String Methods Enhancements

Continuing from v0.16.0, the following enhancements make string operations easier and more consistent with standard
python string operations.
Added StringMethods (.str accessor) to Index (GH9068)
The .str accessor is now available for both Series and Index.

1.3. v0.16.1 (May 11, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [26]: idx = Index([' jack', 'jill ', ' jesse ', 'frank'])
In [27]: idx.str.strip()
Out[27]: Index([u'jack', u'jill', u'jesse', u'frank'], dtype='object')

One special case for the .str accessor on Index is that if a string method returns bool, the .str accessor
will return a np.array instead of a boolean Index (GH8875). This enables the following expression to work
naturally:
In [28]: idx = Index(['a1', 'a2', 'b1', 'b2'])
In [29]: s = Series(range(4), index=idx)
In [30]: s
Out[30]:
a1
0
a2
1
b1
2
b2
3
dtype: int64
In [31]: idx.str.startswith('a')
Out[31]: array([ True, True, False, False], dtype=bool)
In [32]: s[s.index.str.startswith('a')]
Out[32]:
a1
0
a2
1
dtype: int64

The following new methods are accesible via .str accessor to apply the function to each values. (GH9766,
GH9773, GH10031, GH10045, GH10052)
capitalize()
index()

swapcase()
rindex()

Methods
normalize()
translate()

partition()

rpartition()

split now takes expand keyword to specify whether to expand dimensionality. return_type is deprecated. (GH9847)
In [33]: s = Series(['a,b', 'a,c', 'b,c'])
# return Series
In [34]: s.str.split(',')
Out[34]:
0
[a, b]
1
[a, c]
2
[b, c]
dtype: object
# return DataFrame
In [35]: s.str.split(',', expand=True)
Out[35]:
0 1
0 a b
1 a c
2 b c
In [36]: idx = Index(['a,b', 'a,c', 'b,c'])

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

# return Index
In [37]: idx.str.split(',')
Out[37]: Index([[u'a', u'b'], [u'a', u'c'], [u'b', u'c']], dtype='object')
# return MultiIndex
In [38]: idx.str.split(',', expand=True)
Out[38]:
MultiIndex(levels=[[u'a', u'b'], [u'b', u'c']],
labels=[[0, 0, 1], [0, 1, 1]])

Improved extract and get_dummies methods for Index.str (GH9980)

Other Enhancements
BusinessHour offset is now supported, which represents business hours starting from 09:00 - 17:00 on
BusinessDay by default. See Here for details. (GH7905)
In [39]: from pandas.tseries.offsets import BusinessHour
In [40]: Timestamp('2014-08-01 09:00') + BusinessHour()
Out[40]: Timestamp('2014-08-01 10:00:00')
In [41]: Timestamp('2014-08-01 07:00') + BusinessHour()
Out[41]: Timestamp('2014-08-01 10:00:00')
In [42]: Timestamp('2014-08-01 16:30') + BusinessHour()
Out[42]: Timestamp('2014-08-04 09:30:00')

DataFrame.diff now takes an axis parameter that determines the direction of differencing (GH9727)
Allow clip, clip_lower, and clip_upper to accept array-like arguments as thresholds (This is a regression from 0.11.0). These methods now have an axis parameter which determines how the Series or DataFrame
will be aligned with the threshold(s). (GH6966)
DataFrame.mask() and Series.mask() now support same keywords as where (GH8801)
drop function can now accept errors keyword to suppress ValueError raised when any of label does not
exist in the target data. (GH6736)
In [43]: df = DataFrame(np.random.randn(3, 3), columns=['A', 'B', 'C'])
In [44]: df.drop(['A', 'X'], axis=1, errors='ignore')
Out[44]:
B
C
0 0.991946 0.953324
1 -0.334077 0.002118
2 0.289092 1.321158

Add support for separating years and quarters using dashes, for example 2014-Q1. (GH9688)
Allow conversion of values with dtype datetime64 or timedelta64 to strings using astype(str)
(GH9757)
get_dummies function now accepts sparse keyword. If set to True, the return DataFrame is sparse, e.g.
SparseDataFrame. (GH8823)
Period now accepts datetime64 as value input. (GH9054)
Allow timedelta string conversion when leading zero is missing from time definition, ie 0:00:00 vs 00:00:00.
(GH9570)
1.3. v0.16.1 (May 11, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Allow Panel.shift with axis=items (GH9890)

Trying to write an excel file now raises NotImplementedError if the DataFrame has a MultiIndex
instead of writing a broken Excel file. (GH9794)
Allow Categorical.add_categories to accept Series or np.array. (GH9927)
Add/delete str/dt/cat accessors dynamically from __dir__. (GH9910)
Add normalize as a dt accessor method. (GH10047)
DataFrame and Series now have _constructor_expanddim property as overridable constructor for
one higher dimensionality data. This should be used only when it is really needed, see here
pd.lib.infer_dtype now returns bytes in Python 3 where appropriate. (GH10032)

1.3.2 API changes

When passing in an ax to df.plot( ..., ax=ax), the sharex kwarg will now default to False. The result
is that the visibility of xlabels and xticklabels will not anymore be changed. You have to do that by yourself
for the right axes in your figure or set sharex=True explicitly (but this changes the visible for all axes in the
figure, not only the one which is passed in!). If pandas creates the subplots itself (e.g. no passed in ax kwarg),
then the default is still sharex=True and the visibility changes are applied.
assign() now inserts new columns in alphabetical order. Previously the order was arbitrary. (GH9777)
By default, read_csv and read_table will now try to infer the compression type based on the file extension. Set compression=None to restore the previous behavior (no decompression). (GH9770)
Deprecations
Series.str.splits return_type keyword was removed in favor of expand (GH9847)

1.3.3 Index Representation

The string representation of Index and its sub-classes have now been unified. These will show a single-line display
if there are few values; a wrapped multi-line display for a lot of values (but less than display.max_seq_items;
if lots of items (> display.max_seq_items) will show a truncated display (the head and tail of the data). The
formatting for MultiIndex is unchanges (a multi-line wrapped display). The display width responds to the option
display.max_seq_items, which is defaulted to 100. (GH6482)
Previous Behavior
In [2]: pd.Index(range(4),name='foo')
Out[2]: Int64Index([0, 1, 2, 3], dtype='int64')

In [3]: pd.Index(range(104),name='foo')
Out[3]: Int64Index([0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22,
In [4]: pd.date_range('20130101',periods=4,name='foo',tz='US/Eastern')
Out[4]:
<class 'pandas.tseries.index.DatetimeIndex'>
[2013-01-01 00:00:00-05:00, ..., 2013-01-04 00:00:00-05:00]
Length: 4, Freq: D, Timezone: US/Eastern
In [5]: pd.date_range('20130101',periods=104,name='foo',tz='US/Eastern')
Out[5]:

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

<class 'pandas.tseries.index.DatetimeIndex'>
[2013-01-01 00:00:00-05:00, ..., 2013-04-14 00:00:00-04:00]
Length: 104, Freq: D, Timezone: US/Eastern

New Behavior
In [45]: pd.set_option('display.width', 80)
In [46]: pd.Index(range(4), name='foo')
Out[46]: Int64Index([0, 1, 2, 3], dtype='int64', name=u'foo')
In [47]: pd.Index(range(30), name='foo')
Out[47]:
Int64Index([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16,
17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29],
dtype='int64', name=u'foo')
In [48]: pd.Index(range(104), name='foo')
Out[48]:
Int64Index([ 0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
...
94, 95, 96, 97, 98, 99, 100, 101, 102, 103],
dtype='int64', name=u'foo', length=104)
In [49]: pd.CategoricalIndex(['a','bb','ccc','dddd'], ordered=True, name='foobar')
Out[49]: CategoricalIndex([u'a', u'bb', u'ccc', u'dddd'], categories=[u'a', u'bb', u'ccc', u'dddd'],

In [50]: pd.CategoricalIndex(['a','bb','ccc','dddd']*10, ordered=True, name='foobar')

Out[50]:
CategoricalIndex([u'a', u'bb', u'ccc', u'dddd', u'a', u'bb', u'ccc', u'dddd',
u'a', u'bb', u'ccc', u'dddd', u'a', u'bb', u'ccc', u'dddd',
u'a', u'bb', u'ccc', u'dddd', u'a', u'bb', u'ccc', u'dddd',
u'a', u'bb', u'ccc', u'dddd', u'a', u'bb', u'ccc', u'dddd',
u'a', u'bb', u'ccc', u'dddd', u'a', u'bb', u'ccc', u'dddd'],
categories=[u'a', u'bb', u'ccc', u'dddd'], ordered=True, name=u'foobar', dtype='cate

In [51]: pd.CategoricalIndex(['a','bb','ccc','dddd']*100, ordered=True, name='foobar')

Out[51]:
CategoricalIndex([u'a', u'bb', u'ccc', u'dddd', u'a', u'bb', u'ccc', u'dddd',
u'a', u'bb',
...
u'ccc', u'dddd', u'a', u'bb', u'ccc', u'dddd', u'a', u'bb',
u'ccc', u'dddd'],
categories=[u'a', u'bb', u'ccc', u'dddd'], ordered=True, name=u'foobar', dtype='cate
In [52]: pd.date_range('20130101',periods=4, name='foo', tz='US/Eastern')
Out[52]:
DatetimeIndex(['2013-01-01 00:00:00-05:00', '2013-01-02 00:00:00-05:00',
'2013-01-03 00:00:00-05:00', '2013-01-04 00:00:00-05:00'],
dtype='datetime64[ns, US/Eastern]', name=u'foo', freq='D')
In [53]: pd.date_range('20130101',periods=25, freq='D')
Out[53]:
DatetimeIndex(['2013-01-01', '2013-01-02', '2013-01-03',
'2013-01-05', '2013-01-06', '2013-01-07',
'2013-01-09', '2013-01-10', '2013-01-11',
'2013-01-13', '2013-01-14', '2013-01-15',
'2013-01-17', '2013-01-18', '2013-01-19',

1.3. v0.16.1 (May 11, 2015)

'2013-01-04',
'2013-01-08',
'2013-01-12',
'2013-01-16',
'2013-01-20',

pandas: powerful Python data analysis toolkit, Release 0.17.0

'2013-01-21', '2013-01-22', '2013-01-23', '2013-01-24',

'2013-01-25'],
dtype='datetime64[ns]', freq='D')
In [54]: pd.date_range('20130101',periods=104, name='foo', tz='US/Eastern')
Out[54]:
DatetimeIndex(['2013-01-01 00:00:00-05:00', '2013-01-02 00:00:00-05:00',
'2013-01-03 00:00:00-05:00', '2013-01-04 00:00:00-05:00',
'2013-01-05 00:00:00-05:00', '2013-01-06 00:00:00-05:00',
'2013-01-07 00:00:00-05:00', '2013-01-08 00:00:00-05:00',
'2013-01-09 00:00:00-05:00', '2013-01-10 00:00:00-05:00',
...
'2013-04-05 00:00:00-04:00', '2013-04-06 00:00:00-04:00',
'2013-04-07 00:00:00-04:00', '2013-04-08 00:00:00-04:00',
'2013-04-09 00:00:00-04:00', '2013-04-10 00:00:00-04:00',
'2013-04-11 00:00:00-04:00', '2013-04-12 00:00:00-04:00',
'2013-04-13 00:00:00-04:00', '2013-04-14 00:00:00-04:00'],
dtype='datetime64[ns, US/Eastern]', name=u'foo', length=104, freq='D')

1.3.4 Performance Improvements

Improved csv write performance with mixed dtypes, including datetimes by up to 5x (GH9940)
Improved csv write performance generally by 2x (GH9940)
Improved the performance of pd.lib.max_len_string_array by 5-7x (GH10024)

1.3.5 Bug Fixes

Bug where labels did not appear properly in the legend of DataFrame.plot(), passing label= arguments
works, and Series indices are no longer mutated. (GH9542)
Bug in json serialization causing a segfault when a frame had zero length. (GH9805)
Bug in read_csv where missing trailing delimiters would cause segfault. (GH5664)
Bug in retaining index name on appending (GH9862)
Bug in scatter_matrix draws unexpected axis ticklabels (GH5662)
Fixed bug in StataWriter resulting in changes to input DataFrame upon save (GH9795).
Bug in transform causing length mismatch when null entries were present and a fast aggregator was being
used (GH9697)
Bug in equals causing false negatives when block order differed (GH9330)
Bug in grouping with multiple pd.Grouper where one is non-time based (GH10063)
Bug in read_sql_table error when reading postgres table with timezone (GH7139)
Bug in DataFrame slicing may not retain metadata (GH9776)
Bug where TimdeltaIndex were not properly serialized in fixed HDFStore (GH9635)
Bug with TimedeltaIndex constructor ignoring name when given another TimedeltaIndex as data
(GH10025).
Bug in DataFrameFormatter._get_formatted_index with not applying max_colwidth to the
DataFrame index (GH7856)

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug in .loc with a read-only ndarray data source (GH10043)

Bug in groupby.apply() that would raise if a passed user defined function either returned only None (for
all input). (GH9685)
Always use temporary files in pytables tests (GH9992)
Bug in plotting continuously using secondary_y may not show legend properly. (GH9610, GH9779)
Bug in DataFrame.plot(kind="hist") results in TypeError when DataFrame contains nonnumeric columns (GH9853)
Bug where repeated plotting of DataFrame with a DatetimeIndex may raise TypeError (GH9852)
Bug in setup.py that would allow an incompat cython version to build (GH9827)
Bug in plotting secondary_y incorrectly attaches right_ax property to secondary axes specifying itself
recursively. (GH9861)
Bug in Series.quantile on empty Series of type Datetime or Timedelta (GH9675)
Bug in where causing incorrect results when upcasting was required (GH9731)
Bug in FloatArrayFormatter where decision boundary for displaying small floats in decimal format is
off by one order of magnitude for a given display.precision (GH9764)
Fixed bug where DataFrame.plot() raised an error when both color and style keywords were passed
and there was no color symbol in the style strings (GH9671)
Not showing a DeprecationWarning on combining list-likes with an Index (GH10083)
Bug in read_csv and read_table when using skip_rows parameter if blank lines are present. (GH9832)
Bug in read_csv() interprets index_col=True as 1 (GH9798)
Bug in index equality comparisons using == failing on Index/MultiIndex type incompatibility (GH9785)
Bug in which SparseDataFrame could not take nan as a column name (GH8822)
Bug in to_msgpack and read_msgpack zlib and blosc compression support (GH9783)
Bug GroupBy.size doesnt attach index name properly if grouped by TimeGrouper (GH9925)
Bug causing an exception in slice assignments because length_of_indexer returns wrong results
(GH9995)
Bug in csv parser causing lines with initial whitespace plus one non-space character to be skipped. (GH9710)
Bug in C csv parser causing spurious NaNs when data started with newline followed by whitespace. (GH10022)
Bug causing elements with a null group to spill into the final group when grouping by a Categorical
(GH9603)
Bug where .iloc and .loc behavior is not consistent on empty dataframes (GH9964)
Bug in invalid attribute access on a TimedeltaIndex incorrectly raised ValueError instead of
AttributeError (GH9680)
Bug in unequal comparisons between categorical data and a scalar, which was not in the categories (e.g.
Series(Categorical(list("abc"), ordered=True)) > "d". This returned False for all elements, but now raises a TypeError. Equality comparisons also now return False for == and True for !=.
(GH9848)
Bug in DataFrame __setitem__ when right hand side is a dictionary (GH9874)
Bug in where when dtype is datetime64/timedelta64, but dtype of other is not (GH9804)
Bug in MultiIndex.sortlevel() results in unicode level name breaks (GH9856)

1.3. v0.16.1 (May 11, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug in which groupby.transform incorrectly enforced output dtypes to match input dtypes. (GH9807)
Bug in DataFrame constructor when columns parameter is set, and data is an empty list (GH9939)
Bug in bar plot with log=True raises TypeError if all values are less than 1 (GH9905)
Bug in horizontal bar plot ignores log=True (GH9905)
Bug in PyTables queries that did not return proper results using the index (GH8265, GH9676)
Bug where dividing a dataframe containing values of type Decimal by another Decimal would raise.
(GH9787)
Bug where using DataFrames asfreq would remove the name of the index. (GH9885)
Bug causing extra index point when resample BM/BQ (GH9756)
Changed caching in AbstractHolidayCalendar to be at the instance level rather than at the class level as
the latter can result in unexpected behaviour. (GH9552)
Fixed latex output for multi-indexed dataframes (GH9778)
Bug causing an exception when setting an empty range using DataFrame.loc (GH9596)
Bug in hiding ticklabels with subplots and shared axes when adding a new plot to an existing grid of axes
(GH9158)
Bug in transform and filter when grouping on a categorical variable (GH9921)
Bug in transform when groups are equal in number and dtype to the input index (GH9700)
Google BigQuery connector now imports dependencies on a per-method basis.(GH9713)
Updated BigQuery connector to no longer use deprecated oauth2client.tools.run() (GH8327)
Bug in subclassed DataFrame. It may not return the correct class, when slicing or subsetting it. (GH9632)
Bug in .median() where non-float null values are not handled correctly (GH10040)
Bug in Series.fillna() where it raises if a numerically convertible string is given (GH10092)

1.4 v0.16.0 (March 22, 2015)

This is a major release from 0.15.2 and includes a small number of API changes, several new features, enhancements,
and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this
version.
Highlights include:
DataFrame.assign method, see here
Series.to_coo/from_coo methods to interact with scipy.sparse, see here
Backwards incompatible change
datetime.timedelta, see here

Timedelta

conform

the

.seconds

attribute

with

Changes to the .loc slicing API to conform with the behavior of .ix see here
Changes to the default for ordering in the Categorical constructor, see here
Enhancement to the .str accessor to make string operations easier, see here
The pandas.tools.rplot, pandas.sandbox.qtpandas and pandas.rpy modules are deprecated.
We refer users to external packages like seaborn, pandas-qt and rpy2 for similar or equivalent functionality, see
here

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Check the API Changes and deprecations before updating.

Whats new in v0.16.0
New features
DataFrame Assign
Interaction with scipy.sparse
String Methods Enhancements
Other enhancements
Backwards incompatible API changes
Changes in Timedelta
Indexing Changes
Categorical Changes
Other API Changes
Deprecations
Removal of prior version deprecations/changes
Performance Improvements
Bug Fixes

1.4.1 New features

DataFrame Assign
Inspired by dplyrs mutate verb, DataFrame has a new assign() method. The function signature for assign is
simply **kwargs. The keys are the column names for the new fields, and the values are either a value to be inserted
(for example, a Series or NumPy array), or a function of one argument to be called on the DataFrame. The new
values are inserted, and the entire DataFrame (with all original and new columns) is returned.
In [1]: iris = read_csv('data/iris.data')
In [2]: iris.head()
Out[2]:
SepalLength SepalWidth
0
5.1
3.5
1
4.9
3.0
2
4.7
3.2
3
4.6
3.1
4
5.0
3.6

PetalLength
1.4
1.4
1.3
1.5
1.4

PetalWidth
0.2
0.2
0.2
0.2
0.2

In [3]: iris.assign(sepal_ratio=iris['SepalWidth'] /
Out[3]:
SepalLength SepalWidth PetalLength PetalWidth
0
5.1
3.5
1.4
0.2
1
4.9
3.0
1.4
0.2
2
4.7
3.2
1.3
0.2
3
4.6
3.1
1.5
0.2
4
5.0
3.6
1.4
0.2

Name
Iris-setosa
Iris-setosa
Iris-setosa
Iris-setosa
Iris-setosa
iris['SepalLength']).head()
Name
Iris-setosa
Iris-setosa
Iris-setosa
Iris-setosa
Iris-setosa

sepal_ratio
0.686275
0.612245
0.680851
0.673913
0.720000

Above was an example of inserting a precomputed value. We can also pass in a function to be evalutated.
In [4]: iris.assign(sepal_ratio = lambda x: (x['SepalWidth'] /
...:
x['SepalLength'])).head()
...:
Out[4]:
SepalLength SepalWidth PetalLength PetalWidth
Name sepal_ratio

1.4. v0.16.0 (March 22, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

0
1
2
3
4

5.1
4.9
4.7
4.6
5.0

3.5
3.0
3.2
3.1
3.6

1.4
1.4
1.3
1.5
1.4

0.2
0.2
0.2
0.2
0.2

Iris-setosa
Iris-setosa
Iris-setosa
Iris-setosa
Iris-setosa

0.686275
0.612245
0.680851
0.673913
0.720000

The power of assign comes when used in chains of operations. For example, we can limit the DataFrame to just
those with a Sepal Length greater than 5, calculate the ratio, and plot
In [5]: (iris.query('SepalLength > 5')
...:
.assign(SepalRatio = lambda x: x.SepalWidth / x.SepalLength,
...:
PetalRatio = lambda x: x.PetalWidth / x.PetalLength)
...:
.plot(kind='scatter', x='SepalRatio', y='PetalRatio'))
...:
Out[5]: <matplotlib.axes._subplots.AxesSubplot at 0xa993cbcc>

See the documentation for more. (GH9229)

Interaction with scipy.sparse
Added SparseSeries.to_coo() and SparseSeries.from_coo() methods (GH8048) for converting to
and from scipy.sparse.coo_matrix instances (see here). For example, given a SparseSeries with MultiIndex
we can convert to a scipy.sparse.coo_matrix by specifying the row and column labels as index levels:
In [6]: from numpy import nan
In [7]: s = Series([3.0, nan, 1.0, 3.0, nan, nan])
In [8]: s.index = MultiIndex.from_tuples([(1, 2, 'a', 0),
...:
(1, 2, 'a', 1),
...:
(1, 1, 'b', 0),
...:
(1, 1, 'b', 1),
...:
(2, 1, 'b', 0),
...:
(2, 1, 'b', 1)],
...:
names=['A', 'B', 'C', 'D'])
...:
In [9]: s
Out[9]:
A B C D
1 2 a 0

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

1
NaN
0
1
1
3
2 1 b 0
NaN
1
NaN
dtype: float64
1

# SparseSeries
In [10]: ss = s.to_sparse()
In [11]: ss
Out[11]:
A B C D
1 2 a 0
3
1
NaN
1 b 0
1
1
3
2 1 b 0
NaN
1
NaN
dtype: float64
BlockIndex
Block locations: array([0, 2])
Block lengths: array([1, 2])
In [12]: A, rows, columns = ss.to_coo(row_levels=['A', 'B'],
....:
column_levels=['C', 'D'],
....:
sort_labels=False)
....:
In [13]: A
Out[13]:
<3x4 sparse matrix of type '<type 'numpy.float64'>'
with 3 stored elements in COOrdinate format>
In [14]: A.todense()
Out[14]:
matrix([[ 3., 0., 0.,
[ 0., 0., 1.,
[ 0., 0., 0.,

0.],
3.],
0.]])

In [15]: rows
Out[15]: [(1L, 2L), (1L, 1L), (2L, 1L)]
In [16]: columns
Out[16]: [('a', 0L), ('a', 1L), ('b', 0L), ('b', 1L)]

The from_coo method is a

scipy.sparse.coo_matrix:

convenience

method

for

creating

SparseSeries

from

In [17]: from scipy import sparse

In [18]: A = sparse.coo_matrix(([3.0, 1.0, 2.0], ([1, 0, 0], [0, 2, 3])),
....:
shape=(3, 4))
....:
In [19]: A
Out[19]:
<3x4 sparse matrix of type '<type 'numpy.float64'>'

1.4. v0.16.0 (March 22, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

with 3 stored elements in COOrdinate format>

In [20]: A.todense()
Out[20]:
matrix([[ 0., 0., 1.,
[ 3., 0., 0.,
[ 0., 0., 0.,

2.],
0.],
0.]])

In [21]: ss = SparseSeries.from_coo(A)
In [22]: ss
Out[22]:
0 2
1
3
2
1 0
3
dtype: float64
BlockIndex
Block locations: array([0])
Block lengths: array([3])

String Methods Enhancements

Following new methods are accesible via .str accessor to apply the function to each values. This is intended
to make it more consistent with standard methods on strings. (GH9282, GH9352, GH9386, GH9387, GH9439)
isalnum()
islower()
find()

isalpha()
isupper()
rfind()

Methods
isdigit()
istitle()
ljust()

isdigit()
isnumeric()
rjust()

isspace()
isdecimal()
zfill()

In [23]: s = Series(['abcd', '3456', 'EFGH'])

In [24]: s.str.isalpha()
Out[24]:
0
True
1
False
2
True
dtype: bool
In [25]: s.str.find('ab')
Out[25]:
0
0
1
-1
2
-1
dtype: int64

Series.str.pad() and Series.str.center() now accept fillchar option to specify filling character (GH9352)
In [26]: s = Series(['12', '300', '25'])
In [27]: s.str.pad(5, fillchar='_')
Out[27]:
0
___12
1
__300
2
___25
dtype: object

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Added Series.str.slice_replace(), which previously raised NotImplementedError (GH8888)

In [28]: s = Series(['ABCD', 'EFGH', 'IJK'])
In [29]: s.str.slice_replace(1, 3, 'X')
Out[29]:
0
AXD
1
EXH
2
IX
dtype: object
# replaced with empty char
In [30]: s.str.slice_replace(0, 1)
Out[30]:
0
BCD
1
FGH
2
JK
dtype: object

Other enhancements
Reindex now supports method=nearest for frames or series with a monotonic increasing or decreasing
index (GH9258):
In [31]: df = pd.DataFrame({'x': range(5)})
In [32]: df.reindex([0.2, 1.8, 3.5], method='nearest')
Out[32]:
x
0.2 0
1.8 2
3.5 4

This method is also exposed by the lower level Index.get_indexer and Index.get_loc methods.
The read_excel() functions sheetname argument now accepts a list and None, to get multiple or all sheets
respectively. If more than one sheet is specified, a dictionary is returned. (GH9450)
# Returns the 1st and 4th sheet, as a dictionary of DataFrames.
pd.read_excel('path_to_file.xls',sheetname=['Sheet1',3])

Allow Stata files to be read incrementally with an iterator; support for long strings in Stata files. See the docs
here (GH9493:).
Paths beginning with ~ will now be expanded to begin with the users home directory (GH9066)
Added time interval selection in get_data_yahoo (GH9071)
Added Timestamp.to_datetime64() to complement Timedelta.to_timedelta64() (GH9255)
tseries.frequencies.to_offset() now accepts Timedelta as input (GH9064)
Lag parameter was added to the autocorrelation method of Series, defaults to lag-1 autocorrelation (GH9192)
Timedelta will now accept nanoseconds keyword in constructor (GH9273)
SQL code now safely escapes table and column names (GH8986)
Added auto-complete for Series.str.<tab>, Series.dt.<tab> and Series.cat.<tab>
(GH9322)

1.4. v0.16.0 (March 22, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Index.get_indexer now supports method=pad and method=backfill even for any target array, not just monotonic targets. These methods also work for monotonic decreasing as well as monotonic
increasing indexes (GH9258).
Index.asof now works on all index types (GH9258).
A verbose argument has been augmented in io.read_excel(), defaults to False. Set to True to print
sheet names as they are parsed. (GH9450)
Added days_in_month (compatibility alias daysinmonth) property to Timestamp, DatetimeIndex,
Period, PeriodIndex, and Series.dt (GH9572)
Added decimal option in to_csv to provide formatting for non-. decimal separators (GH781)
Added normalize option for Timestamp to normalized to midnight (GH8794)
Added example for DataFrame import to R using HDF5 file and rhdf5 library. See the documentation for
more (GH9636).

1.4.2 Backwards incompatible API changes

Changes in Timedelta
In v0.15.0 a new scalar type Timedelta was introduced, that is a sub-class of datetime.timedelta. Mentioned
here was a notice of an API change w.r.t. the .seconds accessor. The intent was to provide a user-friendly set of
accessors that give the natural value for that unit, e.g. if you had a Timedelta(1 day, 10:11:12), then
.seconds would return 12. However, this is at odds with the definition of datetime.timedelta, which defines
.seconds as 10 * 3600 + 11 * 60 + 12 == 36672.
So in v0.16.0, we are restoring the API to match that of datetime.timedelta. Further, the component values are
still available through the .components accessor. This affects the .seconds and .microseconds accessors,
and removes the .hours, .minutes, .milliseconds accessors. These changes affect TimedeltaIndex and
the Series .dt accessor as well. (GH9185, GH9139)
Previous Behavior
In [2]: t = pd.Timedelta('1 day, 10:11:12.100123')
In [3]: t.days
Out[3]: 1
In [4]: t.seconds
Out[4]: 12
In [5]: t.microseconds
Out[5]: 123

New Behavior
In [33]: t = pd.Timedelta('1 day, 10:11:12.100123')
In [34]: t.days
Out[34]: 1L
In [35]: t.seconds
Out[35]: 36672L
In [36]: t.microseconds
Out[36]: 100123L

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Using .components allows the full component access

In [37]: t.components
Out[37]: Components(days=1L, hours=10L, minutes=11L, seconds=12L, milliseconds=100L, microseconds=123
In [38]: t.components.seconds
Out[38]: 12L

Indexing Changes
The behavior of a small sub-set of edge cases for using .loc have changed (GH8613). Furthermore we have improved
the content of the error messages that are raised:
Slicing with .loc where the start and/or stop bound is not found in the index is now allowed; this previously
would raise a KeyError. This makes the behavior the same as .ix in this case. This change is only for
slicing, not when indexing with a single label.
In [39]: df = DataFrame(np.random.randn(5,4),
....:
columns=list('ABCD'),
....:
index=date_range('20130101',periods=5))
....:
In [40]: df
Out[40]:
A
B
C
D
2013-01-01 -1.546906 -0.202646 -0.655969 0.193421
2013-01-02 0.553439 1.318152 -0.469305 0.675554
2013-01-03 -1.817027 -0.183109 1.058969 -0.397840
2013-01-04 0.337438 1.047579 1.045938 0.863717
2013-01-05 -0.122092 0.124713 -0.322795 0.841675
In [41]: s = Series(range(5),[-2,-1,1,2,3])
In [42]: s
Out[42]:
-2
0
-1
1
1
2
2
3
3
4
dtype: int64

Previous Behavior
In [4]: df.loc['2013-01-02':'2013-01-10']
KeyError: 'stop bound [2013-01-10] is not in the [index]'
In [6]: s.loc[-10:3]
KeyError: 'start bound [-10] is not the [index]'

New Behavior
In [43]: df.loc['2013-01-02':'2013-01-10']
Out[43]:
A
B
C
D
2013-01-02 0.553439 1.318152 -0.469305 0.675554
2013-01-03 -1.817027 -0.183109 1.058969 -0.397840
2013-01-04 0.337438 1.047579 1.045938 0.863717

1.4. v0.16.0 (March 22, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

2013-01-05 -0.122092

0.124713 -0.322795

0.841675

In [44]: s.loc[-10:3]
Out[44]:
-2
0
-1
1
1
2
2
3
3
4
dtype: int64

Allow slicing with float-like values on an integer index for .ix. Previously this was only enabled for .loc:
Previous Behavior

In [8]: s.ix[-1.0:2]
TypeError: the slice start value [-1.0] is not a proper indexer for this index type (Int64Index)

New Behavior
In [45]: s.ix[-1.0:2]
Out[45]:
-1
1
1
2
2
3
dtype: int64

Provide a useful exception for indexing with an invalid type for that index when using .loc. For example
trying to use .loc on an index of type DatetimeIndex or PeriodIndex or TimedeltaIndex, with an
integer (or a float).
Previous Behavior
In [4]: df.loc[2:3]
KeyError: 'start bound [2] is not the [index]'

New Behavior

In [4]: df.loc[2:3]
TypeError: Cannot do slice indexing on <class 'pandas.tseries.index.DatetimeIndex'> with <type '

Categorical Changes
In prior versions, Categoricals that had an unspecified ordering (meaning no ordered keyword was passed)
were defaulted as ordered Categoricals. Going forward, the ordered keyword in the Categorical constructor
will default to False. Ordering must now be explicit.
Furthermore, previously you could change the ordered attribute of a Categorical by just setting the attribute, e.g. cat.ordered=True; This is now deprecated and you should use cat.as_ordered() or
cat.as_unordered(). These will by default return a new object and not modify the existing object. (GH9347,
GH9190)
Previous Behavior
In [3]: s = Series([0,1,2], dtype='category')
In [4]: s
Out[4]:
0
0

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

1
1
2
2
dtype: category
Categories (3, int64): [0 < 1 < 2]
In [5]: s.cat.ordered
Out[5]: True
In [6]: s.cat.ordered = False
In [7]: s
Out[7]:
0
0
1
1
2
2
dtype: category
Categories (3, int64): [0, 1, 2]

New Behavior
In [46]: s = Series([0,1,2], dtype='category')
In [47]: s
Out[47]:
0
0
1
1
2
2
dtype: category
Categories (3, int64): [0, 1, 2]
In [48]: s.cat.ordered
Out[48]: False
In [49]: s = s.cat.as_ordered()
In [50]: s
Out[50]:
0
0
1
1
2
2
dtype: category
Categories (3, int64): [0 < 1 < 2]
In [51]: s.cat.ordered
Out[51]: True
# you can set in the constructor of the Categorical
In [52]: s = Series(Categorical([0,1,2],ordered=True))
In [53]: s
Out[53]:
0
0
1
1
2
2
dtype: category
Categories (3, int64): [0 < 1 < 2]
In [54]: s.cat.ordered

1.4. v0.16.0 (March 22, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Out[54]: True

For ease of creation of series of categorical data, we have added the ability to pass keywords when calling
.astype(). These are passed directly to the constructor.
In [55]: s = Series(["a","b","c","a"]).astype('category',ordered=True)
In [56]: s
Out[56]:
0
a
1
b
2
c
3
a
dtype: category
Categories (3, object): [a < b < c]
In [57]: s = Series(["a","b","c","a"]).astype('category',categories=list('abcdef'),ordered=False)
In [58]: s
Out[58]:
0
a
1
b
2
c
3
a
dtype: category
Categories (6, object): [a, b, c, d, e, f]

Other API Changes

Index.duplicated now returns np.array(dtype=bool) rather than Index(dtype=object)
containing bool values. (GH8875)
DataFrame.to_json now returns accurate type serialisation for each column for frames of mixed dtype
(GH9037)
Previously data was coerced to a common dtype before serialisation, which for example resulted in integers
being serialised to floats:
In [2]: pd.DataFrame({'i': [1,2], 'f': [3.0, 4.2]}).to_json()
Out[2]: '{"f":{"0":3.0,"1":4.2},"i":{"0":1.0,"1":2.0}}'

Now each column is serialised using its correct dtype:

In [2]: pd.DataFrame({'i': [1,2], 'f': [3.0, 4.2]}).to_json()
Out[2]: '{"f":{"0":3.0,"1":4.2},"i":{"0":1,"1":2}}'

DatetimeIndex, PeriodIndex and TimedeltaIndex.summary now output the same format.

(GH9116)
TimedeltaIndex.freqstr now output the same string format as DatetimeIndex. (GH9116)
Bar and horizontal bar plots no longer add a dashed line along the info axis. The prior style can be achieved
with matplotlibs axhline or axvline methods (GH9088).
Series accessors .dt, .cat and .str now raise AttributeError instead of TypeError if the series
does not contain the appropriate type of data (GH9617). This follows Pythons built-in exception hierarchy
more closely and ensures that tests like hasattr(s, cat) are consistent on both Python 2 and 3.

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Series now supports bitwise operation for integral types (GH9016). Previously even if the input dtypes were
integral, the output dtype was coerced to bool.
Previous Behavior
In [2]: pd.Series([0,1,2,3], list('abcd')) | pd.Series([4,4,4,4], list('abcd'))
Out[2]:
a
True
b
True
c
True
d
True
dtype: bool

New Behavior. If the input dtypes are integral, the output dtype is also integral and the output values are the
result of the bitwise operation.
In [2]: pd.Series([0,1,2,3], list('abcd')) | pd.Series([4,4,4,4], list('abcd'))
Out[2]:
a
4
b
5
c
6
d
7
dtype: int64

During division involving a Series or DataFrame, 0/0 and 0//0 now give np.nan instead of np.inf.
(GH9144, GH8445)
Previous Behavior
In [2]: p = pd.Series([0, 1])
In [3]: p / 0
Out[3]:
0
inf
1
inf
dtype: float64
In [4]: p // 0
Out[4]:
0
inf
1
inf
dtype: float64

New Behavior
In [59]: p = pd.Series([0, 1])
In [60]: p / 0
Out[60]:
0
NaN
1
inf
dtype: float64
In [61]: p // 0
Out[61]:
0
NaN
1
inf
dtype: float64

Series.values_counts and Series.describe for categorical data will now put NaN entries at the

1.4. v0.16.0 (March 22, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

end. (GH9443)
Series.describe for categorical data will now give counts and frequencies of 0, not NaN, for unused
categories (GH9443)
Due to a bug fix, looking up a partial string label with DatetimeIndex.asof now includes values that
match the string, even if they are after the start of the partial string label (GH9258).
Old behavior:
In [4]: pd.to_datetime(['2000-01-31', '2000-02-28']).asof('2000-02')
Out[4]: Timestamp('2000-01-31 00:00:00')

Fixed behavior:
In [62]: pd.to_datetime(['2000-01-31', '2000-02-28']).asof('2000-02')
Out[62]: Timestamp('2000-02-28 00:00:00')

To reproduce the old behavior, simply add more precision to the label (e.g., use 2000-02-01 instead of
2000-02).
Deprecations
The rplot trellis plotting interface is deprecated and will be removed in a future version. We refer to external
packages like seaborn for similar but more refined functionality (GH3445). The documentation includes some
examples how to convert your existing code using rplot to seaborn: rplot docs.
The pandas.sandbox.qtpandas interface is deprecated and will be removed in a future version. We refer
users to the external package pandas-qt. (GH9615)
The pandas.rpy interface is deprecated and will be removed in a future version. Similar functionaility can
be accessed thru the rpy2 project (GH9602)
Adding DatetimeIndex/PeriodIndex to another DatetimeIndex/PeriodIndex is being deprecated as a set-operation. This will be changed to a TypeError in a future version. .union() should be used
for the union set operation. (GH9094)
Subtracting DatetimeIndex/PeriodIndex from another DatetimeIndex/PeriodIndex is being deprecated as a set-operation. This will be changed to an actual numeric subtraction yielding a
TimeDeltaIndex in a future version. .difference() should be used for the differencing set operation.
(GH9094)
Removal of prior version deprecations/changes
DataFrame.pivot_table and crosstabs rows and cols keyword arguments were removed in favor
of index and columns (GH6581)
DataFrame.to_excel and DataFrame.to_csv cols keyword argument was removed in favor of
columns (GH6581)
Removed convert_dummies in favor of get_dummies (GH6581)
Removed value_range in favor of describe (GH6581)

1.4.3 Performance Improvements

Fixed a performance regression for .loc indexing with an array or list-like (GH9126:).
DataFrame.to_json 30x performance improvement for mixed dtype frames. (GH9037)
54

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Performance improvements in MultiIndex.duplicated by working with labels instead of values

(GH9125)
Improved the speed of nunique by calling unique instead of value_counts (GH9129, GH7771)
Performance improvement of up to 10x in DataFrame.count and DataFrame.dropna by taking advantage of homogeneous/heterogeneous dtypes appropriately (GH9136)
Performance improvement of up to 20x in DataFrame.count when using a MultiIndex and the level
keyword argument (GH9163)
Performance and memory usage improvements in merge when key space exceeds int64 bounds (GH9151)
Performance improvements in multi-key groupby (GH9429)
Performance improvements in MultiIndex.sortlevel (GH9445)
Performance and memory usage improvements in DataFrame.duplicated (GH9398)
Cythonized Period (GH9440)
Decreased memory usage on to_hdf (GH9648)

1.4.4 Bug Fixes

Changed .to_html to remove leading/trailing spaces in table body (GH4987)
Fixed issue using read_csv on s3 with Python 3 (GH9452)
Fixed compatibility issue in DatetimeIndex affecting architectures where numpy.int_ defaults to
numpy.int32 (GH8943)
Bug in Panel indexing with an object-like (GH9140)
Bug in the returned Series.dt.components index was reset to the default index (GH9247)
Bug in Categorical.__getitem__/__setitem__ with listlike input getting incorrect results from
indexer coercion (GH9469)
Bug in partial setting with a DatetimeIndex (GH9478)
Bug in groupby for integer and datetime64 columns when applying an aggregator that caused the value to be
changed when the number was sufficiently large (GH9311, GH6620)
Fixed bug in to_sql when mapping a Timestamp object column (datetime column with timezone info) to
the appropriate sqlalchemy type (GH9085).
Fixed bug in to_sql dtype argument not accepting an instantiated SQLAlchemy type (GH9083).
Bug in .loc partial setting with a np.datetime64 (GH9516)
Incorrect dtypes inferred on datetimelike looking Series & on .xs slices (GH9477)
Items in Categorical.unique() (and s.unique() if s is of dtype category) now appear in the order
in which they are originally found, not in sorted order (GH9331). This is now consistent with the behavior for
other dtypes in pandas.
Fixed bug on big endian platforms which produced incorrect results in StataReader (GH8688).
Bug in MultiIndex.has_duplicates when having many levels causes an indexer overflow (GH9075,
GH5873)
Bug in pivot and unstack where nan values would break index alignment (GH4862, GH7401, GH7403,
GH7405, GH7466, GH9497)
Bug in left join on multi-index with sort=True or null values (GH9210).
1.4. v0.16.0 (March 22, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug in MultiIndex where inserting new keys would fail (GH9250).

Bug in groupby when key space exceeds int64 bounds (GH9096).
Bug in unstack with TimedeltaIndex or DatetimeIndex and nulls (GH9491).
Bug in rank where comparing floats with tolerance will cause inconsistent behaviour (GH8365).
Fixed character encoding bug in read_stata and StataReader when loading data from a URL (GH9231).
Bug in adding offsets.Nano to other offets raises TypeError (GH9284)
Bug in DatetimeIndex iteration, related to (GH8890), fixed in (GH9100)
Bugs in resample around DST transitions. This required fixing offset classes so they behave correctly on
DST transitions. (GH5172, GH8744, GH8653, GH9173, GH9468).
Bug in binary operator method (eg .mul()) alignment with integer levels (GH9463).
Bug in boxplot, scatter and hexbin plot may show an unnecessary warning (GH8877)
Bug in subplot with layout kw may show unnecessary warning (GH9464)
Bug in using grouper functions that need passed thru arguments (e.g. axis), when using wrapped function (e.g.
fillna), (GH9221)
DataFrame now properly supports simultaneous copy and dtype arguments in constructor (GH9099)
Bug in read_csv when using skiprows on a file with CR line endings with the c engine. (GH9079)
isnull now detects NaT in PeriodIndex (GH9129)
Bug in groupby .nth() with a multiple column groupby (GH8979)
Bug in DataFrame.where and Series.where coerce numerics to string incorrectly (GH9280)
Bug in DataFrame.where and Series.where raise ValueError when string list-like is passed.
(GH9280)
Accessing Series.str methods on with non-string values now raises TypeError instead of producing
incorrect results (GH9184)
Bug in DatetimeIndex.__contains__ when index has duplicates and is not monotonic increasing
(GH9512)
Fixed division by zero error for Series.kurt() when all values are equal (GH9197)
Fixed issue in the xlsxwriter engine where it added a default General format to cells if no other format
wass applied. This prevented other row or column formatting being applied. (GH9167)
Fixes issue with index_col=False when usecols is also specified in read_csv. (GH9082)
Bug where wide_to_long would modify the input stubnames list (GH9204)
Bug in to_sql not storing float64 values using double precision. (GH9009)
SparseSeries and SparsePanel now accept zero argument constructors (same as their non-sparse counterparts) (GH9272).
Regression in merging Categorical and object dtypes (GH9426)
Bug in read_csv with buffer overflows with certain malformed input files (GH9205)
Bug in groupby MultiIndex with missing pair (GH9049, GH9344)
Fixed bug in Series.groupby where grouping on MultiIndex levels would ignore the sort argument
(GH9444)

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Fix bug in DataFrame.Groupby where sort=False is ignored in the case of Categorical columns.
(GH8868)
Fixed bug with reading CSV files from Amazon S3 on python 3 raising a TypeError (GH9452)
Bug in the Google BigQuery reader where the jobComplete key may be present but False in the query results
(GH8728)
Bug in Series.values_counts with excluding NaN for categorical type Series with dropna=True
(GH9443)
Fixed mising numeric_only option for DataFrame.std/var/sem (GH9201)
Support constructing Panel or Panel4D with scalar data (GH8285)
Series text representation disconnected from max_rows/max_columns (GH7508).
Series number formatting inconsistent when truncated (GH8532).
Previous Behavior
In [2]: pd.options.display.max_rows = 10
In [3]: s = pd.Series([1,1,1,1,1,1,1,1,1,1,0.9999,1,1]*10)
In [4]: s
Out[4]:
0
1
1
1
2
1
...
127
0.9999
128
1.0000
129
1.0000
Length: 130, dtype: float64

New Behavior
0
1
2
3
4
...
125
126
127
128
129
dtype:

1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
0.9999
1.0000
1.0000
float64

A Spurious SettingWithCopy Warning was generated when setting a new item in a frame in some cases
(GH8730)
The following would previously report a SettingWithCopy Warning.
In [1]: df1 = DataFrame({'x': Series(['a','b','c']), 'y': Series(['d','e','f'])})
In [2]: df2 = df1[['x']]
In [3]: df2['y'] = ['g', 'h', 'i']

1.4. v0.16.0 (March 22, 2015)

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.5 v0.15.2 (December 12, 2014)

This is a minor release from 0.15.1 and includes a large number of bug fixes along with several new features, enhancements, and performance improvements. A small number of API changes were necessary to fix existing bugs. We
recommend that all users upgrade to this version.
Enhancements
API Changes
Performance Improvements
Bug Fixes

1.5.1 API changes

Indexing in MultiIndex beyond lex-sort depth is now supported, though a lexically sorted index will have a
better performance. (GH2646)
In [1]: df = pd.DataFrame({'jim':[0, 0, 1, 1],
...:
'joe':['x', 'x', 'z', 'y'],
...:
'jolie':np.random.rand(4)}).set_index(['jim', 'joe'])
...:
In [2]: df
Out[2]:
jolie
jim joe
0
x
x
1
z
y

0.043324
0.561433
0.329668
0.502967

In [3]: df.index.lexsort_depth
Out[3]: 1
# in prior versions this would raise a KeyError
# will now show a PerformanceWarning
In [4]: df.loc[(1, 'z')]
Out[4]:
jolie
jim joe
1
z
0.329668
# lexically sorting
In [5]: df2 = df.sortlevel()
In [6]: df2
Out[6]:
jolie
jim joe
0
x
x
1
y
z

0.043324
0.561433
0.502967
0.329668

In [7]: df2.index.lexsort_depth
Out[7]: 2

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [8]: df2.loc[(1,'z')]
Out[8]:
jolie
jim joe
1
z
0.329668

Bug in unique of Series with category dtype, which returned all categories regardless whether they were
used or not (see GH8559 for the discussion). Previous behaviour was to return all categories:
In [3]: cat = pd.Categorical(['a', 'b', 'a'], categories=['a', 'b', 'c'])
In [4]: cat
Out[4]:
[a, b, a]
Categories (3, object): [a < b < c]
In [5]: cat.unique()
Out[5]: array(['a', 'b', 'c'], dtype=object)

Now, only the categories that do effectively occur in the array are returned:
In [9]: cat = pd.Categorical(['a', 'b', 'a'], categories=['a', 'b', 'c'])
In [10]: cat.unique()
Out[10]:
[a, b]
Categories (2, object): [a, b]

Series.all and Series.any now support the level and skipna parameters. Series.all,
Series.any, Index.all, and Index.any no longer support the out and keepdims parameters, which
existed for compatibility with ndarray. Various index types no longer support the all and any aggregation
functions and will now raise TypeError. (GH8302).
Allow equality comparisons of Series with a categorical dtype and object dtype; previously these would raise
TypeError (GH8938)
Bug in NDFrame: conflicting attribute/column names now behave consistently between getting and setting.
Previously, when both a column and attribute named y existed, data.y would return the attribute, while
data.y = z would update the column (GH8994)
In [11]: data = pd.DataFrame({'x':[1, 2, 3]})
In [12]: data.y = 2
In [13]: data['y'] = [2, 4, 6]
In [14]: data
Out[14]:
x y
0 1 2
1 2 4
2 3 6
# this assignment was inconsistent
In [15]: data.y = 5

Old behavior:

1.5. v0.15.2 (December 12, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [6]: data.y
Out[6]: 2
In [7]: data['y'].values
Out[7]: array([5, 5, 5])

New behavior:
In [16]: data.y
Out[16]: 5
In [17]: data['y'].values
Out[17]: array([2, 4, 6], dtype=int64)

Timestamp(now) is now equivalent to Timestamp.now() in that it returns the local time rather than
UTC. Also, Timestamp(today) is now equivalent to Timestamp.today() and both have tz as a
possible argument. (GH9000)
Fix negative step support for label-based slices (GH8753)
Old behavior:
In [1]: s = pd.Series(np.arange(3), ['a', 'b', 'c'])
Out[1]:
a
0
b
1
c
2
dtype: int64
In [2]: s.loc['c':'a':-1]
Out[2]:
c
2
dtype: int64

New behavior:
In [18]: s = pd.Series(np.arange(3), ['a', 'b', 'c'])
In [19]: s.loc['c':'a':-1]
Out[19]:
c
2
b
1
a
0
dtype: int32

1.5.2 Enhancements
Categorical enhancements:
Added ability to export Categorical data to Stata (GH8633). See here for limitations of categorical variables
exported to Stata data files.
Added flag order_categoricals to StataReader and read_stata to select whether to order imported categorical data (GH8836). See here for more information on importing categorical variables from Stata
data files.
Added ability to export Categorical data to to/from HDF5 (GH7621). Queries work the same as if it was an
object array. However, the category dtyped data is stored in a more efficient manner. See here for an
example and caveats w.r.t. prior versions of pandas.
60

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Added support for searchsorted() on Categorical class (GH8420).

Other enhancements:
Added the ability to specify the SQL type of columns when writing a DataFrame to a database (GH8778). For
example, specifying to use the sqlalchemy String type instead of the default Text type for string columns:
from sqlalchemy.types import String
data.to_sql('data_dtype', engine, dtype={'Col_1': String})

Series.all and Series.any now support the level and skipna parameters (GH8302):
In [20]: s = pd.Series([False, True, False], index=[0, 0, 1])
In [21]: s.any(level=0)
Out[21]:
0
True
1
False
dtype: bool

Panel now supports the all and any aggregation functions. (GH8302):
In [22]: p = pd.Panel(np.random.rand(2, 5, 4) > 0.1)
In [23]: p.all()
Out[23]:
0
1
0
True False
1
True
True
2
True
True
3 False
True

Added support for utcfromtimestamp(), fromtimestamp(), and combine() on Timestamp class

(GH5351).
Added Google Analytics (pandas.io.ga) basic documentation (GH8835). See here.
Timedelta arithmetic returns NotImplemented in unknown cases, allowing extensions by custom classes
(GH8813).
Timedelta now supports arithemtic with numpy.ndarray objects of the appropriate dtype (numpy 1.8 or
newer only) (GH8884).
Added Timedelta.to_timedelta64() method to the public API (GH8884).
Added gbq.generate_bq_schema() function to the gbq module (GH8325).
Series now works with map objects the same way as generators (GH8909).
Added context manager to HDFStore for automatic closing (GH8791).
to_datetime gains an exact keyword to allow for a format to not require an exact match for a provided format string (if its False). exact defaults to True (meaning that exact matching is still the default) (GH8904)
Added axvlines boolean option to parallel_coordinates plot function, determines whether vertical lines will
be printed, default is True
Added ability to read table footers to read_html (GH8552)
to_sql now infers datatypes of non-NA values for columns that contain NA values and have dtype object
(GH8778).

1.5. v0.15.2 (December 12, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.5.3 Performance
Reduce memory usage when skiprows is an integer in read_csv (GH8681)
Performance boost for to_datetime conversions with a passed format=, and the exact=False
(GH8904)

1.5.4 Bug Fixes

Bug in concat of Series with category dtype which were coercing to object. (GH8641)
Bug in Timestamp-Timestamp not returning a Timedelta type and datelike-datelike ops with timezones
(GH8865)
Made consistent a timezone mismatch exception (either tz operated with None or incompatible timezone), will
now return TypeError rather than ValueError (a couple of edge cases only), (GH8865)
Bug in using a pd.Grouper(key=...) with no level/axis or level only (GH8795, GH8866)
Report a TypeError when invalid/no paramaters are passed in a groupby (GH8015)
Bug in packaging pandas with py2app/cx_Freeze (GH8602, GH8831)
Bug in groupby signatures that didnt include *args or **kwargs (GH8733).
io.data.Options now raises RemoteDataError when no expiry dates are available from Yahoo and
when it receives no data from Yahoo (GH8761), (GH8783).
Unclear error message in csv parsing when passing dtype and names and the parsed data is a different data type
(GH8833)
Bug in slicing a multi-index with an empty list and at least one boolean indexer (GH8781)
io.data.Options now raises RemoteDataError when no expiry dates are available from Yahoo
(GH8761).
Timedelta kwargs may now be numpy ints and floats (GH8757).
Fixed several outstanding bugs for Timedelta arithmetic and comparisons (GH8813, GH5963, GH5436).
sql_schema now generates dialect appropriate CREATE TABLE statements (GH8697)
slice string method now takes step into account (GH8754)
Bug in BlockManager where setting values with different type would break block integrity (GH8850)
Bug in DatetimeIndex when using time object as key (GH8667)
Bug in merge where how=left and sort=False would not preserve left frame order (GH7331)
Bug in MultiIndex.reindex where reindexing at level would not reorder labels (GH4088)
Bug in certain operations with dateutil timezones, manifesting with dateutil 2.3 (GH8639)
Regression in DatetimeIndex iteration with a Fixed/Local offset timezone (GH8890)
Bug in to_datetime when parsing a nanoseconds using the %f format (GH8989)
io.data.Options now raises RemoteDataError when no expiry dates are available from Yahoo and
when it receives no data from Yahoo (GH8761), (GH8783).
Fix: The font size was only set on x axis if vertical or the y axis if horizontal. (GH8765)
Fixed division by 0 when reading big csv files in python 3 (GH8621)
Bug in outputing a Multindex with to_html,index=False which would add an extra column (GH8452)

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Imported categorical variables from Stata files retain the ordinal information in the underlying data (GH8836).
Defined .size attribute across NDFrame objects to provide compat with numpy >= 1.9.1; buggy with
np.array_split (GH8846)
Skip testing of histogram plots for matplotlib <= 1.2 (GH8648).
Bug where get_data_google returned object dtypes (GH3995)
Bug in DataFrame.stack(..., dropna=False) when the DataFrames columns is a MultiIndex
whose labels do not reference all its levels. (GH8844)
Bug in that Option context applied on __enter__ (GH8514)
Bug in resample that causes a ValueError when resampling across multiple days and the last offset is not calculated from the start of the range (GH8683)
Bug where DataFrame.plot(kind=scatter) fails when checking if an np.array is in the DataFrame
(GH8852)
Bug in pd.infer_freq/DataFrame.inferred_freq that prevented proper sub-daily frequency inference when the index contained DST days (GH8772).
Bug where index name was still used when plotting a series with use_index=False (GH8558).
Bugs when trying to stack multiple columns, when some (or all) of the level names are numbers (GH8584).
Bug in MultiIndex where __contains__ returns wrong result if index is not lexically sorted or unique
(GH7724)
BUG CSV: fix problem with trailing whitespace in skipped rows, (GH8679), (GH8661), (GH8983)
Regression in Timestamp does not parse Z zone designator for UTC (GH8771)
Bug in StataWriter the produces writes strings with 244 characters irrespective of actual size (GH8969)
Fixed ValueError raised by cummin/cummax when datetime64 Series contains NaT. (GH8965)
Bug in Datareader returns object dtype if there are missing values (GH8980)
Bug in plotting if sharex was enabled and index was a timeseries, would show labels on multiple axes (GH3964).
Bug where passing a unit to the TimedeltaIndex constructor applied the to nano-second conversion twice.
(GH9011).
Bug in plotting of a period-like array (GH9012)

1.6 v0.15.1 (November 9, 2014)

This is a minor bug-fix release from 0.15.0 and includes a small number of API changes, several new features, enhancements, and performance improvements along with a large number of bug fixes. We recommend that all users
upgrade to this version.
Enhancements
API Changes
Bug Fixes

1.6. v0.15.1 (November 9, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.6.1 API changes

s.dt.hour and other .dt accessors will now return np.nan for missing values (rather than previously -1),
(GH8689)
In [1]: s = Series(date_range('20130101',periods=5,freq='D'))
In [2]: s.iloc[2] = np.nan
In [3]: s
Out[3]:
0
2013-01-01
1
2013-01-02
2
NaT
3
2013-01-04
4
2013-01-05
dtype: datetime64[ns]

previous behavior:
In [6]: s.dt.hour
Out[6]:
0
0
1
0
2
-1
3
0
4
0
dtype: int64

current behavior:
In [4]: s.dt.hour
Out[4]:
0
0
1
0
2
NaN
3
0
4
0
dtype: float64

groupby with as_index=False will not add erroneous extra columns to result (GH8582):
In [5]: np.random.seed(2718281)
In [6]: df = pd.DataFrame(np.random.randint(0, 100, (10, 2)),
...:
columns=['jim', 'joe'])
...:
In [7]: df.head()
Out[7]:
jim joe
0
61
81
1
96
49
2
55
65
3
72
51
4
77
12
In [8]: ts = pd.Series(5 * np.random.randint(0, 3, 10))

previous behavior:
64

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [4]: df.groupby(ts, as_index=False).max()

Out[4]:
NaN jim joe
0
0
72
83
1
5
77
84
2
10
96
65

current behavior:
In [9]: df.groupby(ts, as_index=False).max()
Out[9]:
jim joe
0
72
83
1
77
84
2
96
65

groupby will not erroneously exclude columns if the column name conflics with the grouper name (GH8112):
In [10]: df = pd.DataFrame({'jim': range(5), 'joe': range(5, 10)})
In [11]: df
Out[11]:
jim joe
0
0
5
1
1
6
2
2
7
3
3
8
4
4
9
In [12]: gr = df.groupby(df['jim'] < 2)

previous behavior (excludes 1st column from output):

In [4]: gr.apply(sum)
Out[4]:
joe
jim
False
24
True
11

current behavior:
In [13]: gr.apply(sum)
Out[13]:
jim joe
jim
False
9
24
True
1
11

Support for slicing with monotonic decreasing indexes, even if start or stop is not found in the index
(GH7860):
In [14]: s = pd.Series(['a', 'b', 'c', 'd'], [4, 3, 2, 1])
In [15]: s
Out[15]:
4
a
3
b
2
c

1.6. v0.15.1 (November 9, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

1
d
dtype: object

previous behavior:
In [8]: s.loc[3.5:1.5]
KeyError: 3.5

current behavior:
In [16]: s.loc[3.5:1.5]
Out[16]:
3
b
2
c
dtype: object

io.data.Options has been fixed for a change in the format of the Yahoo Options page (GH8612),
(GH8741)
Note: As a result of a change in Yahoos option page layout, when an expiry date is given, Options methods
now return data for a single expiry date. Previously, methods returned all data for the selected month.
The month and year parameters have been undeprecated and can be used to get all options data for a given
month.
If an expiry date that is not valid is given, data for the next expiry after the given date is returned.
Option data frames are now saved on the instance as callsYYMMDD or putsYYMMDD. Previously they were
saved as callsMMYY and putsMMYY. The next expiry is saved as calls and puts.
New features:
The expiry parameter can now be a single date or a list-like object containing dates.
A new property expiry_dates was added, which returns all available expiry dates.
Current behavior:
In [17]: from pandas.io.data import Options
In [18]: aapl = Options('aapl','yahoo')
In [19]: aapl.get_call_data().iloc[0:5,0:1]
Out[19]:
Last
Strike
60
65
70
75
79

Expiry
2015-10-16
2015-10-16
2015-10-16
2015-10-16
2015-10-16

Type
call
call
call
call
call

Symbol
AAPL151016C00060000
AAPL151016C00065000
AAPL151016C00070000
AAPL151016C00075000
AAPL151016C00079000

51.00
47.84
38.68
36.65
30.80

In [20]: aapl.expiry_dates
Out[20]:
[datetime.date(2015, 10, 16),
datetime.date(2015, 10, 23),
datetime.date(2015, 10, 30),
datetime.date(2015, 11, 6),
datetime.date(2015, 11, 13),
datetime.date(2015, 11, 20),

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

datetime.date(2015,
datetime.date(2016,
datetime.date(2016,
datetime.date(2016,
datetime.date(2016,
datetime.date(2017,
datetime.date(2018,

12, 18),
1, 15),
4, 15),
6, 17),
7, 15),
1, 20),
1, 19)]

In [21]: aapl.get_near_stock_price(expiry=aapl.expiry_dates[0:3]).iloc[0:5,0:1]
Out[21]:
Last
Strike Expiry
Type Symbol
111
2015-10-23 call AAPL151023C00111000 2.44
2015-10-30 call AAPL151030C00111000 4.45
112
2015-10-16 call AAPL151016C00112000 1.13
2015-10-23 call AAPL151023C00112000 1.85
2015-10-30 call AAPL151030C00112000 3.85

See the Options documentation in Remote Data

pandas now also registers the datetime64 dtype in matplotlibs units registry to plot such values as datetimes. This is activated once pandas is imported. In previous versions, plotting an array of datetime64
values will have resulted in plotted integer values. To keep the previous behaviour, you can do del
matplotlib.units.registry[np.datetime64] (GH8614).

1.6.2 Enhancements
concat permits a wider variety of iterables of pandas objects to be passed as the first parameter (GH8645):
In [22]: from collections import deque
In [23]: df1 = pd.DataFrame([1, 2, 3])
In [24]: df2 = pd.DataFrame([4, 5, 6])

previous behavior:

In [7]: pd.concat(deque((df1, df2)))

TypeError: first argument must be a list-like of pandas objects, you passed an object of type "d

current behavior:
In [25]: pd.concat(deque((df1, df2)))
Out[25]:
0
0 1
1 2
2 3
0 4
1 5
2 6

Represent MultiIndex labels with a dtype that utilizes memory based on the level size. In prior versions,
the memory usage was a constant 8 bytes per element in each level. In addition, in prior versions, the reported
memory usage was incorrect as it didnt show the usage for the memory occupied by the underling data array.
(GH8456)

1.6. v0.15.1 (November 9, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [26]: dfi = DataFrame(1,index=pd.MultiIndex.from_product([['a'],range(1000)]),columns=['A'])

previous behavior:
# this was underreported in prior versions
In [1]: dfi.memory_usage(index=True)
Out[1]:
Index
8000 # took about 24008 bytes in < 0.15.1
A
8000
dtype: int64

current behavior:
In [27]: dfi.memory_usage(index=True)
Out[27]:
Index
11020
A
8000
dtype: int64

Added Index properties is_monotonic_increasing and is_monotonic_decreasing (GH8680).

Added option to select columns when importing Stata files (GH7935)
Qualify memory usage in DataFrame.info() by adding + if it is a lower bound (GH8578)
Raise errors in certain aggregation cases where an argument such as numeric_only is not handled (GH8592).
Added support for 3-character ISO and non-standard country codes in io.wb.download() (GH8482)
World Bank data requests now will warn/raise based on an errors argument, as well as a list of hard-coded
country codes and the World Banks JSON response. In prior versions, the error messages didnt look at the
World Banks JSON response. Problem-inducing input were simply dropped prior to the request. The issue was
that many good countries were cropped in the hard-coded approach. All countries will work now, but some bad
countries will raise exceptions because some edge cases break the entire response. (GH8482)
Added option to Series.str.split() to return a DataFrame rather than a Series (GH8428)
Added option to df.info(null_counts=None|True|False) to override the default display options
and force showing of the null-counts (GH8701)

1.6.3 Bug Fixes

Bug in unpickling of a CustomBusinessDay object (GH8591)
Bug in coercing Categorical to a records array, e.g. df.to_records() (GH8626)
Bug in Categorical not created properly with Series.to_frame() (GH8626)
Bug in coercing in astype of a Categorical of a passed pd.Categorical (this now raises TypeError
correctly), (GH8626)
Bug in cut/qcut when using Series and retbins=True (GH8589)
Bug in writing Categorical columns to an SQL database with to_sql (GH8624).
Bug in comparing Categorical of datetime raising when being compared to a scalar datetime (GH8687)
Bug in selecting from a Categorical with .iloc (GH8623)
Bug in groupby-transform with a Categorical (GH8623)
Bug in duplicated/drop_duplicates with a Categorical (GH8623)

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug in Categorical reflected comparison operator raising if the first argument was a numpy array scalar
(e.g. np.int64) (GH8658)
Bug in Panel indexing with a list-like (GH8710)
Compat issue is DataFrame.dtypes when options.mode.use_inf_as_null is True (GH8722)
Bug in read_csv, dialect parameter would not take a string (:issue: 8703)
Bug in slicing a multi-index level with an empty-list (GH8737)
Bug in numeric index operations of add/sub with Float/Index Index with numpy arrays (GH8608)
Bug in setitem with empty indexer and unwanted coercion of dtypes (GH8669)
Bug in ix/loc block splitting on setitem (manifests with integer-like dtypes, e.g. datetime64) (GH8607)
Bug when doing label based indexing with integers not found in the index for non-unique but monotonic indexes
(GH8680).
Bug when indexing a Float64Index with np.nan on numpy 1.7 (GH8980).
Fix shape attribute for MultiIndex (GH8609)
Bug in GroupBy where a name conflict between the grouper and columns would break groupby operations
(GH7115, GH8112)
Fixed a bug where plotting a column y and specifying a label would mutate the index name of the original
DataFrame (GH8494)
Fix regression in plotting of a DatetimeIndex directly with matplotlib (GH8614).
Bug in date_range where partially-specified dates would incorporate current date (GH6961)
Bug in Setting by indexer to a scalar value with a mixed-dtype Panel4d was failing (GH8702)
Bug where DataReaders would fail if one of the symbols passed was invalid. Now returns data for valid
symbols and np.nan for invalid (GH8494)
Bug in get_quote_yahoo that wouldnt allow non-float return values (GH5229).

1.7 v0.15.0 (October 18, 2014)

This is a major release from 0.14.1 and includes a small number of API changes, several new features, enhancements,
and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this
version.
Warning: pandas >= 0.15.0 will no longer support compatibility with NumPy versions < 1.7.0. If you want to
use the latest versions of pandas, please upgrade to NumPy >= 1.7.0 (GH7711)
Highlights include:
The Categorical type was integrated as a first-class pandas type, see here
New scalar type Timedelta, and a new index type TimedeltaIndex, see here
New datetimelike properties accessor .dt for Series, see Datetimelike Properties
New DataFrame default display for df.info() to include memory usage, see Memory Usage
read_csv will now by default ignore blank lines when parsing, see here
API change in using Indexes in set operations, see here
Enhancements in the handling of timezones, see here
1.7. v0.15.0 (October 18, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

A lot of improvements to the rolling and expanding moment funtions, see here
Internal refactoring of the Index class to no longer sub-class ndarray, see Internal Refactoring
dropping support for PyTables less than version 3.0.0, and numexpr less than version 2.1 (GH7990)
Split indexing documentation into Indexing and Selecting Data and MultiIndex / Advanced Indexing
Split out string methods documentation into Working with Text Data
Check the API Changes and deprecations before updating
Other Enhancements
Performance Improvements
Bug Fixes
Warning: In 0.15.0 Index has internally been refactored to no longer sub-class ndarray but instead subclass
PandasObject, similarly to the rest of the pandas objects. This change allows very easy sub-classing and
creation of new index types. This should be a transparent change with only very limited API implications (See the
Internal Refactoring)
Warning: The refactorings in Categorical changed the two argument constructor from codes/labels and
levels to values and levels (now called categories). This can lead to subtle bugs. If you use Categorical
directly, please audit your code before updating to this pandas version and change it to use the from_codes()
constructor. See more on Categorical here

1.7.1 New features

Categoricals in Series/DataFrame
Categorical can now be included in Series and DataFrames and gained new methods to manipulate. Thanks to Jan
Schulz for much of this API/implementation. (GH3943, GH5313, GH5314, GH7444, GH7839, GH7848, GH7864,
GH7914, GH7768, GH8006, GH3678, GH8075, GH8076, GH8143, GH8453, GH8518).
For full docs, see the categorical introduction and the API documentation.
In [1]: df = DataFrame({"id":[1,2,3,4,5,6], "raw_grade":['a', 'b', 'b', 'a', 'a', 'e']})
In [2]: df["grade"] = df["raw_grade"].astype("category")
In [3]: df["grade"]
Out[3]:
0
a
1
b
2
b
3
a
4
a
5
e
Name: grade, dtype: category
Categories (3, object): [a, b, e]
# Rename the categories
In [4]: df["grade"].cat.categories = ["very good", "good", "very bad"]

# Reorder the categories and simultaneously add the missing categories

In [5]: df["grade"] = df["grade"].cat.set_categories(["very bad", "bad", "medium", "good", "very good

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [6]: df["grade"]
Out[6]:
0
very good
1
good
2
good
3
very good
4
very good
5
very bad
Name: grade, dtype: category
Categories (5, object): [very bad, bad, medium, good, very good]
In [7]: df.sort("grade")
Out[7]:
id raw_grade
grade
5
6
e
very bad
1
2
b
good
2
3
b
good
0
1
a very good
3
4
a very good
4
5
a very good
In [8]: df.groupby("grade").size()
Out[8]:
grade
very bad
1
bad
0
medium
0
good
2
very good
3
dtype: int64

pandas.core.group_agg and pandas.core.factor_agg were removed. As an alternative, construct a dataframe and use df.groupby(<group>).agg(<func>).
Supplying codes/labels and levels to the Categorical constructor is not supported anymore. Supplying
two arguments to the constructor is now interpreted as values and levels (now called categories). Please
change your code to use the from_codes() constructor.
The Categorical.labels attribute was renamed to Categorical.codes and is read only. If you want
to manipulate codes, please use one of the API methods on Categoricals.
The Categorical.levels attribute is renamed to Categorical.categories.
TimedeltaIndex/Scalar
We introduce a new scalar type Timedelta, which is a subclass of datetime.timedelta, and behaves in a
similar manner, but allows compatibility with np.timedelta64 types as well as a host of custom representation,
parsing, and attributes. This type is very similar to how Timestamp works for datetimes. It is a nice-API box for
the type. See the docs. (GH3009, GH4533, GH8209, GH8187, GH8190, GH7869, GH7661, GH8345, GH8471)

1.7. v0.15.0 (October 18, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Warning: Timedelta scalars (and TimedeltaIndex) component fields are not the same as the component
fields on a datetime.timedelta object. For example, .seconds on a datetime.timedelta object
returns the total number of seconds combined between hours, minutes and seconds. In contrast, the pandas
Timedelta breaks out hours, minutes, microseconds and nanoseconds separately.
# Timedelta accessor
In [9]: tds = Timedelta('31 days 5 min 3 sec')
In [10]: tds.minutes
Out[10]: 5L
In [11]: tds.seconds
Out[11]: 3L
# datetime.timedelta accessor
# this is 5 minutes * 60 + 3 seconds
In [12]: tds.to_pytimedelta().seconds
Out[12]: 303

Note: this is no longer true starting from v0.16.0, where full compatibility with datetime.timedelta is
introduced. See the 0.16.0 whatsnew entry
Warning: Prior to 0.15.0 pd.to_timedelta would return a Series for list-like/Series input, and a
np.timedelta64 for scalar input. It will now return a TimedeltaIndex for list-like input, Series for
Series input, and Timedelta for scalar input.
The arguments to pd.to_timedelta are now (arg,unit=ns,box=True,coerce=False), previously were (arg,box=True,unit=ns) as these are more logical.
Consruct a scalar
In [9]: Timedelta('1 days 06:05:01.00003')
Out[9]: Timedelta('1 days 06:05:01.000030')
In [10]: Timedelta('15.5us')
Out[10]: Timedelta('0 days 00:00:00.000015')
In [11]: Timedelta('1 hour 15.5us')
Out[11]: Timedelta('0 days 01:00:00.000015')
# negative Timedeltas have this string repr
# to be more consistent with datetime.timedelta conventions
In [12]: Timedelta('-1us')
Out[12]: Timedelta('-1 days +23:59:59.999999')
# a NaT
In [13]: Timedelta('nan')
Out[13]: NaT

Access fields for a Timedelta

In [14]: td = Timedelta('1 hour 3m 15.5us')
In [15]: td.seconds
Out[15]: 3780L
In [16]: td.microseconds
Out[16]: 15L

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [17]: td.nanoseconds
Out[17]: 500L

Construct a TimedeltaIndex
In [18]: TimedeltaIndex(['1 days','1 days, 00:00:05',
....:
np.timedelta64(2,'D'),timedelta(days=2,seconds=2)])
....:
Out[18]:
TimedeltaIndex(['1 days 00:00:00', '1 days 00:00:05', '2 days 00:00:00',
'2 days 00:00:02'],
dtype='timedelta64[ns]', freq=None)

Constructing a TimedeltaIndex with a regular range

In [19]: timedelta_range('1 days',periods=5,freq='D')
Out[19]: TimedeltaIndex(['1 days', '2 days', '3 days', '4 days', '5 days'], dtype='timedelta64[ns]',
In [20]: timedelta_range(start='1 days',end='2 days',freq='30T')
Out[20]:
TimedeltaIndex(['1 days 00:00:00', '1 days 00:30:00', '1 days 01:00:00',
'1 days 01:30:00', '1 days 02:00:00', '1 days 02:30:00',
'1 days 03:00:00', '1 days 03:30:00', '1 days 04:00:00',
'1 days 04:30:00', '1 days 05:00:00', '1 days 05:30:00',
'1 days 06:00:00', '1 days 06:30:00', '1 days 07:00:00',
'1 days 07:30:00', '1 days 08:00:00', '1 days 08:30:00',
'1 days 09:00:00', '1 days 09:30:00', '1 days 10:00:00',
'1 days 10:30:00', '1 days 11:00:00', '1 days 11:30:00',
'1 days 12:00:00', '1 days 12:30:00', '1 days 13:00:00',
'1 days 13:30:00', '1 days 14:00:00', '1 days 14:30:00',
'1 days 15:00:00', '1 days 15:30:00', '1 days 16:00:00',
'1 days 16:30:00', '1 days 17:00:00', '1 days 17:30:00',
'1 days 18:00:00', '1 days 18:30:00', '1 days 19:00:00',
'1 days 19:30:00', '1 days 20:00:00', '1 days 20:30:00',
'1 days 21:00:00', '1 days 21:30:00', '1 days 22:00:00',
'1 days 22:30:00', '1 days 23:00:00', '1 days 23:30:00',
'2 days 00:00:00'],
dtype='timedelta64[ns]', freq='30T')

You can now use a TimedeltaIndex as the index of a pandas object

In [21]: s = Series(np.arange(5),
....:
index=timedelta_range('1 days',periods=5,freq='s'))
....:
In [22]: s
Out[22]:
1 days 00:00:00
0
1 days 00:00:01
1
1 days 00:00:02
2
1 days 00:00:03
3
1 days 00:00:04
4
Freq: S, dtype: int32

You can select with partial string selections

In [23]: s['1 day 00:00:02']
Out[23]: 2

1.7. v0.15.0 (October 18, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [24]: s['1 day':'1 day 00:00:02']

Out[24]:
1 days 00:00:00
0
1 days 00:00:01
1
1 days 00:00:02
2
Freq: S, dtype: int32

Finally, the combination of TimedeltaIndex with DatetimeIndex allow certain combination operations that
are NaT preserving:
In [25]: tdi = TimedeltaIndex(['1 days',pd.NaT,'2 days'])
In [26]: tdi.tolist()
Out[26]: [Timedelta('1 days 00:00:00'), NaT, Timedelta('2 days 00:00:00')]
In [27]: dti = date_range('20130101',periods=3)
In [28]: dti.tolist()
Out[28]:
[Timestamp('2013-01-01 00:00:00', offset='D'),
Timestamp('2013-01-02 00:00:00', offset='D'),
Timestamp('2013-01-03 00:00:00', offset='D')]
In [29]: (dti + tdi).tolist()
Out[29]: [Timestamp('2013-01-02 00:00:00'), NaT, Timestamp('2013-01-05 00:00:00')]
In [30]: (dti - tdi).tolist()
Out[30]: [Timestamp('2012-12-31 00:00:00'), NaT, Timestamp('2013-01-01 00:00:00')]

iteration of a Series e.g. list(Series(...)) of timedelta64[ns] would prior to v0.15.0 return

np.timedelta64 for each element. These will now be wrapped in Timedelta.
Memory Usage
Implemented methods to find memory usage of a DataFrame. See the FAQ for more. (GH6852).
A new display option display.memory_usage (see Options and Settings) sets the default behavior of the
memory_usage argument in the df.info() method. By default display.memory_usage is True.
In [31]: dtypes = ['int64', 'float64', 'datetime64[ns]', 'timedelta64[ns]',
....:
'complex128', 'object', 'bool']
....:
In [32]: n = 5000
In [33]: data = dict([ (t, np.random.randint(100, size=n).astype(t))
....:
for t in dtypes])
....:
In [34]: df = DataFrame(data)
In [35]: df['categorical'] = df['object'].astype('category')
In [36]: df.info()
<class 'pandas.core.frame.DataFrame'>
Int64Index: 5000 entries, 0 to 4999
Data columns (total 8 columns):
bool
5000 non-null bool

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

complex128
5000 non-null complex128
datetime64[ns]
5000 non-null datetime64[ns]
float64
5000 non-null float64
int64
5000 non-null int64
object
5000 non-null object
timedelta64[ns]
5000 non-null timedelta64[ns]
categorical
5000 non-null category
dtypes: bool(1), category(1), complex128(1), datetime64[ns](1), float64(1), int64(1), object(1), time
memory usage: 303.5+ KB

Additionally memory_usage() is an available method for a dataframe object which returns the memory usage of
each column.
In [37]: df.memory_usage(index=True)
Out[37]:
Index
40000
bool
5000
complex128
80000
datetime64[ns]
40000
float64
40000
int64
40000
object
20000
timedelta64[ns]
40000
categorical
5800
dtype: int64

.dt accessor
Series has gained an accessor to succinctly return datetime like properties for the values of the Series, if its a
datetime/period like Series. (GH7207) This will return a Series, indexed like the existing Series. See the docs
# datetime
In [38]: s = Series(date_range('20130101 09:10:12',periods=4))
In [39]: s
Out[39]:
0
2013-01-01 09:10:12
1
2013-01-02 09:10:12
2
2013-01-03 09:10:12
3
2013-01-04 09:10:12
dtype: datetime64[ns]
In [40]: s.dt.hour
Out[40]:
0
9
1
9
2
9
3
9
dtype: int64
In [41]: s.dt.second
Out[41]:
0
12
1
12
2
12
3
12
dtype: int64

1.7. v0.15.0 (October 18, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [42]: s.dt.day
Out[42]:
0
1
1
2
2
3
3
4
dtype: int64
In [43]: s.dt.freq
Out[43]: <Day>

This enables nice expressions like this:

In [44]: s[s.dt.day==2]
Out[44]:
1
2013-01-02 09:10:12
dtype: datetime64[ns]

You can easily produce tz aware transformations:

In [45]: stz = s.dt.tz_localize('US/Eastern')
In [46]: stz
Out[46]:
0
2013-01-01 09:10:12-05:00
1
2013-01-02 09:10:12-05:00
2
2013-01-03 09:10:12-05:00
3
2013-01-04 09:10:12-05:00
dtype: datetime64[ns, US/Eastern]
In [47]: stz.dt.tz
Out[47]: <DstTzInfo 'US/Eastern' LMT-1 day, 19:04:00 STD>

You can also chain these types of operations:

In [48]: s.dt.tz_localize('UTC').dt.tz_convert('US/Eastern')
Out[48]:
0
2013-01-01 04:10:12-05:00
1
2013-01-02 04:10:12-05:00
2
2013-01-03 04:10:12-05:00
3
2013-01-04 04:10:12-05:00
dtype: datetime64[ns, US/Eastern]

The .dt accessor works for period and timedelta dtypes.

# period
In [49]: s = Series(period_range('20130101',periods=4,freq='D'))
In [50]: s
Out[50]:
0
2013-01-01
1
2013-01-02
2
2013-01-03
3
2013-01-04
dtype: object
In [51]: s.dt.year
Out[51]:
0
2013

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

1
2013
2
2013
3
2013
dtype: int64
In [52]: s.dt.day
Out[52]:
0
1
1
2
2
3
3
4
dtype: int64
# timedelta
In [53]: s = Series(timedelta_range('1 day 00:00:05',periods=4,freq='s'))
In [54]: s
Out[54]:
0
1 days 00:00:05
1
1 days 00:00:06
2
1 days 00:00:07
3
1 days 00:00:08
dtype: timedelta64[ns]
In [55]: s.dt.days
Out[55]:
0
1
1
1
2
1
3
1
dtype: int64
In [56]: s.dt.seconds
Out[56]:
0
5
1
6
2
7
3
8
dtype: int64
In [57]: s.dt.components
Out[57]:
days hours minutes seconds
0
1
0
0
5
1
1
0
0
6
2
1
0
0
7
3
1
0
0
8

milliseconds
0
0
0
0

microseconds
0
0
0
0

nanoseconds
0
0
0
0

Timezone handling improvements

tz_localize(None) for tz-aware Timestamp and DatetimeIndex now removes timezone holding
local time, previously this resulted in Exception or TypeError (GH7812)
In [58]: ts = Timestamp('2014-08-01 09:00', tz='US/Eastern')
In [59]: ts
Out[59]: Timestamp('2014-08-01 09:00:00-0400', tz='US/Eastern')

1.7. v0.15.0 (October 18, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [60]: ts.tz_localize(None)
Out[60]: Timestamp('2014-08-01 09:00:00')
In [61]: didx = DatetimeIndex(start='2014-08-01 09:00', freq='H', periods=10, tz='US/Eastern')
In [62]: didx
Out[62]:
DatetimeIndex(['2014-08-01 09:00:00-04:00', '2014-08-01 10:00:00-04:00',
'2014-08-01 11:00:00-04:00', '2014-08-01 12:00:00-04:00',
'2014-08-01 13:00:00-04:00', '2014-08-01 14:00:00-04:00',
'2014-08-01 15:00:00-04:00', '2014-08-01 16:00:00-04:00',
'2014-08-01 17:00:00-04:00', '2014-08-01 18:00:00-04:00'],
dtype='datetime64[ns, US/Eastern]', freq='H')
In [63]: didx.tz_localize(None)
Out[63]:
DatetimeIndex(['2014-08-01 09:00:00',
'2014-08-01 11:00:00',
'2014-08-01 13:00:00',
'2014-08-01 15:00:00',
'2014-08-01 17:00:00',
dtype='datetime64[ns]',

'2014-08-01
'2014-08-01
'2014-08-01
'2014-08-01
'2014-08-01
freq='H')

10:00:00',
12:00:00',
14:00:00',
16:00:00',
18:00:00'],

tz_localize now accepts the ambiguous keyword which allows for passing an array of bools indicating
whether the date belongs in DST or not, NaT for setting transition times to NaT, infer for inferring DST/nonDST, and raise (default) for an AmbiguousTimeError to be raised. See the docs for more details (GH7943)
DataFrame.tz_localize and DataFrame.tz_convert now accepts an optional level argument
for localizing a specific level of a MultiIndex (GH7846)
Timestamp.tz_localize and Timestamp.tz_convert now raise TypeError in error cases, rather
than Exception (GH8025)
a timeseries/index localized to UTC when inserted into a Series/DataFrame will preserve the UTC timezone
(rather than being a naive datetime64[ns]) as object dtype (GH8411)
Timestamp.__repr__ displays dateutil.tz.tzoffset info (GH7907)
Rolling/Expanding Moments improvements
rolling_min(), rolling_max(), rolling_cov(), and rolling_corr() now return objects with
all NaN when len(arg) < min_periods <= window rather than raising. (This makes all rolling functions consistent in this behavior). (GH7766)
Prior to 0.15.0
In [64]: s = Series([10, 11, 12, 13])
In [15]: rolling_min(s, window=10, min_periods=5)
ValueError: min_periods (5) must be <= window (4)

New behavior
In [65]: rolling_min(s, window=10, min_periods=5)
Out[65]:
0
NaN
1
NaN
2
NaN

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

3
NaN
dtype: float64

rolling_max(), rolling_min(), rolling_sum(), rolling_mean(), rolling_median(),

rolling_std(),
rolling_var(),
rolling_skew(),
rolling_kurt(),
rolling_quantile(), rolling_cov(), rolling_corr(), rolling_corr_pairwise(),
rolling_window(), and rolling_apply() with center=True previously would return a result of
the same structure as the input arg with NaN in the final (window-1)/2 entries.
Now the final (window-1)/2 entries of the result are calculated as if the input arg were followed by
(window-1)/2 NaN values (or with shrinking windows, in the case of rolling_apply()). (GH7925,
GH8269)
Prior behavior (note final value is NaN):
In [7]: rolling_sum(Series(range(4)), window=3, min_periods=0, center=True)
Out[7]:
0
1
1
3
2
6
3
NaN
dtype: float64

New behavior (note final value is 5 = sum([2, 3, NaN])):

In [66]: rolling_sum(Series(range(4)), window=3, min_periods=0, center=True)
Out[66]:
0
1
1
3
2
6
3
5
dtype: float64

rolling_window() now normalizes the weights properly in rolling mean mode (mean=True) so that the
calculated weighted means (e.g. triang, gaussian) are distributed about the same means as those calculated
without weighting (i.e. boxcar). See the note on normalization for further details. (GH7618)
In [67]: s = Series([10.5, 8.8, 11.4, 9.7, 9.3])

Behavior prior to 0.15.0:

In [39]: rolling_window(s, window=3, win_type='triang', center=True)
Out[39]:
0
NaN
1
6.583333
2
6.883333
3
6.683333
4
NaN
dtype: float64

New behavior
In [68]: rolling_window(s, window=3, win_type='triang', center=True)
Out[68]:
0
NaN
1
9.875
2
10.325
3
10.025
4
NaN
dtype: float64

1.7. v0.15.0 (October 18, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Removed center argument from all expanding_ functions (see list), as the results produced when
center=True did not make much sense. (GH7925)
Added optional ddof argument to expanding_cov() and rolling_cov(). The default value of 1 is
backwards-compatible. (GH8279)
Documented the ddof argument to expanding_var(), expanding_std(), rolling_var(), and
rolling_std(). These functions support of a ddof argument (with a default value of 1) was previously
undocumented. (GH8064)
ewma(), ewmstd(), ewmvol(), ewmvar(), ewmcov(), and ewmcorr() now interpret min_periods
in the same manner that the rolling_*() and expanding_*() functions do: a given result entry will be
NaN if the (expanding, in this case) window does not contain at least min_periods values. The previous
behavior was to set to NaN the min_periods entries starting with the first non- NaN value. (GH7977)
Prior behavior (note values start at index 2, which is min_periods after index 0 (the index of the first nonempty value)):
In [69]: s

= Series([1, None, None, None, 2, 3])

In [51]: ewma(s, com=3., min_periods=2)

Out[51]:
0
NaN
1
NaN
2
1.000000
3
1.000000
4
1.571429
5
2.189189
dtype: float64

New behavior (note values start at index 4, the location of the 2nd (since min_periods=2) non-empty value):
In [70]: ewma(s, com=3., min_periods=2)
Out[70]:
0
NaN
1
NaN
2
NaN
3
NaN
4
1.759644
5
2.383784
dtype: float64

ewmstd(), ewmvol(), ewmvar(), ewmcov(), and ewmcorr() now have an optional adjust argument, just like ewma() does, affecting how the weights are calculated. The default value of adjust is True,
which is backwards-compatible. See Exponentially weighted moment functions for details. (GH7911)
ewma(), ewmstd(), ewmvol(), ewmvar(), ewmcov(), and ewmcorr() now have an optional
ignore_na argument. When ignore_na=False (the default), missing values are taken into account in
the weights calculation. When ignore_na=True (which reproduces the pre-0.15.0 behavior), missing values
are ignored in the weights calculation. (GH7543)
In [71]: ewma(Series([None, 1., 8.]), com=2.)
Out[71]:
0
NaN
1
1.0
2
5.2
dtype: float64
In [72]: ewma(Series([1., None, 8.]), com=2., ignore_na=True)
Out[72]:

# pre-0.15.0 behavior

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

0
1.0
1
1.0
2
5.2
dtype: float64
In [73]: ewma(Series([1., None, 8.]), com=2., ignore_na=False)
Out[73]:
0
1.000000
1
1.000000
2
5.846154
dtype: float64

# new default

Warning: By default (ignore_na=False) the ewm*() functions weights calculation in the presence
of missing values is different than in pre-0.15.0 versions. To reproduce the pre-0.15.0 calculation of weights
in the presence of missing values one must specify explicitly ignore_na=True.
Bug in expanding_cov(), expanding_corr(), rolling_cov(), rolling_cor(), ewmcov(),
and ewmcorr() returning results with columns sorted by name and producing an error for non-unique
columns; now handles non-unique columns and returns columns in original order (except for the case of two
DataFrames with pairwise=False, where behavior is unchanged) (GH7542)
Bug in rolling_count() and expanding_*() functions unnecessarily producing error message for
zero-length data (GH8056)
Bug in rolling_apply()
min_periods=1 (GH8080)

and

expanding_apply()

interpreting

min_periods=0

Bug in expanding_std() and expanding_var() for a single value producing a confusing error message
(GH7900)
Bug in rolling_std() and rolling_var() for a single value producing 0 rather than NaN (GH7900)
Bug in ewmstd(), ewmvol(), ewmvar(), and ewmcov() calculation of de-biasing factors when
bias=False (the default). Previously an incorrect constant factor was used, based on adjust=True,
ignore_na=True, and an infinite number of observations. Now a different factor is used for each entry,
based on the actual weights (analogous to the usual N/(N-1) factor). In particular, for a single point a value of
NaN is returned when bias=False, whereas previously a value of (approximately) 0 was returned.
For example, consider the following pre-0.15.0 results for ewmvar(..., bias=False), and the corresponding debiasing factors:
In [74]: s = Series([1., 2., 0., 4.])
In [89]: ewmvar(s, com=2., bias=False)
Out[89]:
0
-2.775558e-16
1
3.000000e-01
2
9.556787e-01
3
3.585799e+00
dtype: float64
In [90]: ewmvar(s, com=2., bias=False) / ewmvar(s, com=2., bias=True)
Out[90]:
0
1.25
1
1.25
2
1.25
3
1.25
dtype: float64

1.7. v0.15.0 (October 18, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Note that entry 0 is approximately 0, and the debiasing factors are a constant 1.25. By comparison, the following
0.15.0 results have a NaN for entry 0, and the debiasing factors are decreasing (towards 1.25):
In [75]: ewmvar(s, com=2., bias=False)
Out[75]:
0
NaN
1
0.500000
2
1.210526
3
4.089069
dtype: float64
In [76]: ewmvar(s, com=2., bias=False) / ewmvar(s, com=2., bias=True)
Out[76]:
0
NaN
1
2.083333
2
1.583333
3
1.425439
dtype: float64

See Exponentially weighted moment functions for details. (GH7912)

Improvements in the sql io module
Added support for a chunksize parameter to to_sql function. This allows DataFrame to be written in
chunks and avoid packet-size overflow errors (GH8062).
Added support for a chunksize parameter to read_sql function. Specifying this argument will return an
iterator through chunks of the query result (GH2908).
Added support for writing datetime.date and datetime.time object columns with to_sql (GH6932).
Added support for specifying a schema to read from/write to with read_sql_table and to_sql
(GH7441, GH7952). For example:
df.to_sql('table', engine, schema='other_schema')
pd.read_sql_table('table', engine, schema='other_schema')

Added support for writing NaN values with to_sql (GH2754).

Added support for writing datetime64 columns with to_sql for all database flavors (GH7103).

1.7.2 Backwards incompatible API changes

Breaking changes
API changes related to Categorical (see here for more details):
The Categorical constructor with two arguments changed from codes/labels and levels to values and
levels (now called categories). This can lead to subtle bugs. If you use Categorical directly, please audit
your code by changing it to use the from_codes() constructor.
An old function call like (prior to 0.15.0):
pd.Categorical([0,1,0,2,1], levels=['a', 'b', 'c'])

will have to adapted to the following to keep the same behaviour:

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [2]: pd.Categorical.from_codes([0,1,0,2,1], categories=['a', 'b', 'c'])

Out[2]:
[a, b, a, c, b]
Categories (3, object): [a, b, c]

API changes related to the introduction of the Timedelta scalar (see above for more details):
Prior to 0.15.0 to_timedelta() would return a Series for list-like/Series input, and a
np.timedelta64 for scalar input. It will now return a TimedeltaIndex for list-like input, Series
for Series input, and Timedelta for scalar input.
For API changes related to the rolling and expanding functions, see detailed overview above.
Other notable API changes:
Consistency when indexing with .loc and a list-like indexer when no values are found.
In [77]: df = DataFrame([['a'],['b']],index=[1,2])
In [78]: df
Out[78]:
0
1 a
2 b

In prior versions there was a difference in these two constructs:

df.loc[[3]] would return a frame reindexed by 3 (with all np.nan values)
df.loc[[3],:] would raise KeyError.
Both will now raise a KeyError. The rule is that at least 1 indexer must be found when using a list-like and
.loc (GH7999)
Furthermore in prior versions these were also different:
df.loc[[1,3]] would return a frame reindexed by [1,3]
df.loc[[1,3],:] would raise KeyError.
Both will now return a frame reindex by [1,3]. E.g.
In [79]: df.loc[[1,3]]
Out[79]:
0
1
a
3 NaN
In [80]: df.loc[[1,3],:]
Out[80]:
0
1
a
3 NaN

This can also be seen in multi-axis indexing with a Panel.

In [81]: p = Panel(np.arange(2*3*4).reshape(2,3,4),
....:
items=['ItemA','ItemB'],
....:
major_axis=[1,2,3],
....:
minor_axis=['A','B','C','D'])
....:
In [82]: p

1.7. v0.15.0 (October 18, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Out[82]:
<class 'pandas.core.panel.Panel'>
Dimensions: 2 (items) x 3 (major_axis) x 4 (minor_axis)
Items axis: ItemA to ItemB
Major_axis axis: 1 to 3
Minor_axis axis: A to D

The following would raise KeyError prior to 0.15.0:

In [83]: p.loc[['ItemA','ItemD'],:,'D']
Out[83]:
ItemA ItemD
1
3
NaN
2
7
NaN
3
11
NaN

Furthermore, .loc will raise If no values are found in a multi-index with a list-like indexer:
In [84]: s = Series(np.arange(3,dtype='int64'),
....:
index=MultiIndex.from_product([['A'],['foo','bar','baz']],
....:
names=['one','two'])
....:
).sortlevel()
....:
In [85]: s
Out[85]:
one two
A
bar
1
baz
2
foo
0
dtype: int64
In [86]: try:
....:
s.loc[['D']]
....: except KeyError as e:
....:
print("KeyError: " + str(e))
....:
KeyError: 'cannot index a multi-index axis with these keys'

Assigning values to None now considers the dtype when choosing an empty value (GH7941).
Previously, assigning to None in numeric containers changed the dtype to object (or errored, depending on the
call). It now uses NaN:
In [87]: s = Series([1, 2, 3])
In [88]: s.loc[0] = None
In [89]: s
Out[89]:
0
NaN
1
2
2
3
dtype: float64

NaT is now used similarly for datetime containers.

For object containers, we now preserve None values (previously these were converted to NaN values).

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [90]: s = Series(["a", "b", "c"])

In [91]: s.loc[0] = None
In [92]: s
Out[92]:
0
None
1
b
2
c
dtype: object

To insert a NaN, you must explicitly use np.nan. See the docs.
In prior versions, updating a pandas object inplace would not reflect in other python references to this object.
(GH8511, GH5104)
In [93]: s = Series([1, 2, 3])
In [94]: s2 = s
In [95]: s += 1.5

Behavior prior to v0.15.0

# the original object
In [5]: s
Out[5]:
0
2.5
1
3.5
2
4.5
dtype: float64

# a reference to the original object

In [7]: s2
Out[7]:
0
1
1
2
2
3
dtype: int64

This is now the correct behavior

# the original object
In [96]: s
Out[96]:
0
2.5
1
3.5
2
4.5
dtype: float64
# a reference to the original object
In [97]: s2
Out[97]:
0
2.5
1
3.5
2
4.5
dtype: float64

Made both the C-based and Python engines for read_csv and read_table ignore empty lines in input as well as
1.7. v0.15.0 (October 18, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

whitespace-filled lines, as long as sep is not whitespace. This is an API change that can be controlled by the
keyword parameter skip_blank_lines. See the docs (GH4466)
A timeseries/index localized to UTC when inserted into a Series/DataFrame will preserve the UTC timezone
and inserted as object dtype rather than being converted to a naive datetime64[ns] (GH8411).
Bug in passing a DatetimeIndex with a timezone that was not being retained in DataFrame construction
from a dict (GH7822)
In prior versions this would drop the timezone, now it retains the timezone, but gives a column of object
dtype:
In [98]: i = date_range('1/1/2011', periods=3, freq='10s', tz = 'US/Eastern')
In [99]: i
Out[99]:
DatetimeIndex(['2011-01-01 00:00:00-05:00', '2011-01-01 00:00:10-05:00',
'2011-01-01 00:00:20-05:00'],
dtype='datetime64[ns, US/Eastern]', freq='10S')
In [100]: df = DataFrame( {'a' : i } )
In [101]: df
Out[101]:
a
0 2011-01-01 00:00:00-05:00
1 2011-01-01 00:00:10-05:00
2 2011-01-01 00:00:20-05:00
In [102]: df.dtypes
Out[102]:
a
datetime64[ns, US/Eastern]
dtype: object

Previously this would have yielded a column of datetime64 dtype, but without timezone info.
The behaviour of assigning a column to an existing dataframe as df[a] = i remains unchanged (this already
returned an object column with a timezone).
When passing multiple levels to stack(), it will now raise a ValueError when the levels arent all level
names or all level numbers (GH7660). See Reshaping by stacking and unstacking.
Raise a ValueError in df.to_hdf with fixed format, if df has non-unique columns as the resulting file
will be broken (GH7761)
SettingWithCopy raise/warnings (according to the option mode.chained_assignment) will now be
issued when setting a value on a sliced mixed-dtype DataFrame using chained-assignment. (GH7845, GH7950)
In [1]: df = DataFrame(np.arange(0,9), columns=['count'])
In [2]: df['group'] = 'b'
In [3]: df.iloc[0:5]['group'] = 'a'
/usr/local/bin/ipython:1: SettingWithCopyWarning:
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.h

merge, DataFrame.merge, and ordered_merge now return the same type as the left argument
(GH7737).

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Previously an enlargement with a mixed-dtype frame would act unlike .append which will preserve dtypes
(related GH2578, GH8176):
In [103]: df = DataFrame([[True, 1],[False, 2]],
.....:
columns=["female","fitness"])
.....:
In [104]: df
Out[104]:
female fitness
0
True
1
1 False
2
In [105]: df.dtypes
Out[105]:
female
bool
fitness
int64
dtype: object
# dtypes are now preserved
In [106]: df.loc[2] = df.loc[1]
In [107]: df
Out[107]:
female fitness
0
True
1
1 False
2
2 False
2
In [108]: df.dtypes
Out[108]:
female
bool
fitness
int64
dtype: object

Series.to_csv() now returns

DataFrame.to_csv() (GH8215).

string

when

path=None,

matching

the

behaviour

read_hdf now raises IOError when a file that doesnt exist is passed in. Previously, a new, empty file was
created, and a KeyError raised (GH7715).
DataFrame.info() now ends its output with a newline character (GH8114)
Concatenating no objects will now raise a ValueError rather than a bare Exception.
Merge errors will now be sub-classes of ValueError rather than raw Exception (GH8501)
DataFrame.plot and Series.plot keywords are now have consistent orders (GH8037)
Internal Refactoring
In 0.15.0 Index has internally been refactored to no longer sub-class ndarray but instead subclass
PandasObject, similarly to the rest of the pandas objects. This change allows very easy sub-classing and creation of new index types. This should be a transparent change with only very limited API implications (GH5080,
GH7439, GH7796, GH8024, GH8367, GH7997, GH8522):
you may need to unpickle pandas version < 0.15.0 pickles using pd.read_pickle rather than
pickle.load. See pickle docs

1.7. v0.15.0 (October 18, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

when plotting with a PeriodIndex, the matplotlib internal axes will now be arrays of Period rather than a
PeriodIndex (this is similar to how a DatetimeIndex passes arrays of datetimes now)
MultiIndexes will now raise similary to other pandas objects w.r.t. truth testing, see here (GH7897).
When plotting a DatetimeIndex directly with matplotlibs plot function, the axis labels will no longer be formatted as dates but as integers (the internal representation of a datetime64). UPDATE This is fixed in 0.15.1,
see here.
Deprecations
The attributes Categorical labels and levels attributes are deprecated and renamed to codes and
categories.
The outtype argument to pd.DataFrame.to_dict has been deprecated in favor of orient. (GH7840)
The convert_dummies method has been deprecated in favor of get_dummies (GH8140)
The infer_dst argument in tz_localize will be deprecated in favor of ambiguous to allow for more
flexibility in dealing with DST transitions. Replace infer_dst=True with ambiguous=infer for the
same behavior (GH7943). See the docs for more details.
The top-level pd.value_range has been deprecated and can be replaced by .describe() (GH8481)
The Index set operations + and - were deprecated in order to provide these for numeric type operations on
certain index types. + can be replaced by .union() or |, and - by .difference(). Further the method
name Index.diff() is deprecated and can be replaced by Index.difference() (GH8226)
# +
Index(['a','b','c']) + Index(['b','c','d'])
# should be replaced by
Index(['a','b','c']).union(Index(['b','c','d']))
# Index(['a','b','c']) - Index(['b','c','d'])
# should be replaced by
Index(['a','b','c']).difference(Index(['b','c','d']))

The infer_types argument to read_html() now has no effect and is deprecated (GH7762, GH7032).
Removal of prior version deprecations/changes
Remove DataFrame.delevel method in favor of DataFrame.reset_index

1.7.3 Enhancements
Enhancements in the importing/exporting of Stata files:
Added support for bool, uint8, uint16 and uint32 datatypes in to_stata (GH7097, GH7365)
Added conversion option when importing Stata files (GH8527)
DataFrame.to_stata and StataWriter check string length for compatibility with limitations imposed
in dta files where fixed-width strings must contain 244 or fewer characters. Attempting to write Stata dta files
with strings longer than 244 characters raises a ValueError. (GH7858)

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

read_stata and StataReader can import missing data information into a DataFrame by setting
the argument convert_missing to True. When using this options, missing values are returned as
StataMissingValue objects and columns containing missing values have object data type. (GH8045)
Enhancements in the plotting functions:
Added layout keyword to DataFrame.plot. You can pass a tuple of (rows, columns), one of which
can be -1 to automatically infer (GH6667, GH8071).
Allow to pass multiple axes to DataFrame.plot, hist and boxplot (GH5353, GH6970, GH7069)
Added support for c, colormap
kind=scatter (GH7780)

and

colorbar

arguments

for

DataFrame.plot

with

Histogram from DataFrame.plot with kind=hist (GH7809), See the docs.

Boxplot from DataFrame.plot with kind=box (GH7998), See the docs.
Other:
read_csv now has a keyword parameter float_precision which specifies which floating-point converter
the C engine should use during parsing, see here (GH8002, GH8044)
Added searchsorted method to Series objects (GH7447)
describe() on mixed-types DataFrames is more flexible. Type-based column filtering is now possible via
the include/exclude arguments. See the docs (GH8164).
In [109]: df = DataFrame({'catA': ['foo', 'foo', 'bar'] * 8,
.....:
'catB': ['a', 'b', 'c', 'd'] * 6,
.....:
'numC': np.arange(24),
.....:
'numD': np.arange(24.) + .5})
.....:
In [110]: df.describe(include=["object"])
Out[110]:
catA catB
count
24
24
unique
2
4
top
foo
d
freq
16
6
In [111]: df.describe(include=["number", "object"], exclude=["float"])
Out[111]:
catA catB
numC
count
24
24 24.000000
unique
2
4
NaN
top
foo
d
NaN
freq
16
6
NaN
mean
NaN NaN 11.500000
std
NaN NaN
7.071068
min
NaN NaN
0.000000
25%
NaN NaN
5.750000
50%
NaN NaN 11.500000
75%
NaN NaN 17.250000
max
NaN NaN 23.000000

Requesting all columns is possible with the shorthand all

In [112]: df.describe(include='all')
Out[112]:
catA catB
numC
numD

1.7. v0.15.0 (October 18, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

count
unique
top
freq
mean
std
min
25%
50%
75%
max

24
2
foo
16
NaN
NaN
NaN
NaN
NaN
NaN
NaN

24
4
d
6
NaN
NaN
NaN
NaN
NaN
NaN
NaN

24.000000
NaN
NaN
NaN
11.500000
7.071068
0.000000
5.750000
11.500000
17.250000
23.000000

24.000000
NaN
NaN
NaN
12.000000
7.071068
0.500000
6.250000
12.000000
17.750000
23.500000

Without those arguments, describe will behave as before, including only numerical columns or, if none are,
only categorical columns. See also the docs
Added split as an option to the orient argument in pd.DataFrame.to_dict. (GH7840)
The get_dummies method can now be used on DataFrames. By default only catagorical columns are encoded
as 0s and 1s, while other columns are left untouched.
In [113]: df = DataFrame({'A': ['a', 'b', 'a'], 'B': ['c', 'c', 'b'],
.....:
'C': [1, 2, 3]})
.....:
In [114]: pd.get_dummies(df)
Out[114]:
C A_a A_b B_b B_c
0 1
1
0
0
1
1 2
0
1
0
1
2 3
1
0
1
0

PeriodIndex supports resolution as the same as DatetimeIndex (GH7708)

pandas.tseries.holiday has added support for additional holidays and ways to observe holidays
(GH7070)
pandas.tseries.holiday.Holiday now supports a list of offsets in Python3 (GH7070)
pandas.tseries.holiday.Holiday now supports a days_of_week parameter (GH7070)
GroupBy.nth() now supports selecting multiple nth values (GH7910)
In [115]: business_dates = date_range(start='4/1/2014', end='6/30/2014', freq='B')
In [116]: df = DataFrame(1, index=business_dates, columns=['a', 'b'])
# get the first, 4th, and last date index for each month
In [117]: df.groupby((df.index.year, df.index.month)).nth([0, 3, -1])
Out[117]:
a b
2014-04-01 1 1
2014-04-04 1 1
2014-04-30 1 1
2014-05-01 1 1
2014-05-06 1 1
2014-05-30 1 1
2014-06-02 1 1
2014-06-05 1 1
2014-06-30 1 1

Period and PeriodIndex supports addition/subtraction with timedelta-likes (GH7966)

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

If Period freq is D, H, T, S, L, U, N, Timedelta-like can be added if the result can have same freq. Otherwise,
only the same offsets can be added.
In [118]: idx = pd.period_range('2014-07-01 09:00', periods=5, freq='H')
In [119]: idx
Out[119]:
PeriodIndex(['2014-07-01 09:00', '2014-07-01 10:00', '2014-07-01 11:00',
'2014-07-01 12:00', '2014-07-01 13:00'],
dtype='int64', freq='H')
In [120]: idx + pd.offsets.Hour(2)
Out[120]:
PeriodIndex(['2014-07-01 11:00', '2014-07-01 12:00', '2014-07-01 13:00',
'2014-07-01 14:00', '2014-07-01 15:00'],
dtype='int64', freq='H')
In [121]: idx + Timedelta('120m')
Out[121]:
PeriodIndex(['2014-07-01 11:00', '2014-07-01 12:00', '2014-07-01 13:00',
'2014-07-01 14:00', '2014-07-01 15:00'],
dtype='int64', freq='H')
In [122]: idx = pd.period_range('2014-07', periods=5, freq='M')

In [123]: idx
Out[123]: PeriodIndex(['2014-07', '2014-08', '2014-09', '2014-10', '2014-11'], dtype='int64', fr

In [124]: idx + pd.offsets.MonthEnd(3)

Out[124]: PeriodIndex(['2014-10', '2014-11', '2014-12', '2015-01', '2015-02'], dtype='int64', fr

Added experimental compatibility with openpyxl for versions >= 2.0. The DataFrame.to_excel method
engine keyword now recognizes openpyxl1 and openpyxl2 which will explicitly require openpyxl v1
and v2 respectively, failing if the requested version is not available. The openpyxl engine is a now a metaengine that automatically uses whichever version of openpyxl is installed. (GH7177)
DataFrame.fillna can now accept a DataFrame as a fill value (GH8377)
Passing multiple levels to stack() will now work when multiple level numbers are passed (GH7660). See
Reshaping by stacking and unstacking.
set_names(), set_labels(), and set_levels() methods now take an optional level keyword argument to all modification of specific level(s) of a MultiIndex. Additionally set_names() now accepts a
scalar string value when operating on an Index or on a specific level of a MultiIndex (GH7792)

In [125]: idx = MultiIndex.from_product([['a'], range(3), list("pqr")], names=['foo', 'bar', 'ba

In [126]: idx.set_names('qux', level=0)
Out[126]:
MultiIndex(levels=[[u'a'], [0, 1, 2], [u'p', u'q', u'r']],
labels=[[0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 1, 1, 1, 2, 2, 2], [0, 1, 2, 0, 1, 2,
names=[u'qux', u'bar', u'baz'])
In [127]: idx.set_names(['qux','baz'], level=[0,1])
Out[127]:
MultiIndex(levels=[[u'a'], [0, 1, 2], [u'p', u'q', u'r']],
labels=[[0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 1, 1, 1, 2, 2, 2], [0, 1, 2, 0, 1, 2,
names=[u'qux', u'baz', u'baz'])
In [128]: idx.set_levels(['a','b','c'], level='bar')

1.7. v0.15.0 (October 18, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Out[128]:
MultiIndex(levels=[[u'a'], [u'a', u'b', u'c'], [u'p', u'q', u'r']],
labels=[[0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 1, 1, 1, 2, 2, 2], [0, 1, 2, 0, 1, 2,
names=[u'foo', u'bar', u'baz'])
In [129]: idx.set_levels([['a','b','c'],[1,2,3]], level=[1,2])
Out[129]:
MultiIndex(levels=[[u'a'], [u'a', u'b', u'c'], [1, 2, 3]],
labels=[[0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 1, 1, 1, 2, 2, 2], [0, 1, 2, 0, 1, 2,
names=[u'foo', u'bar', u'baz'])

Index.isin now supports a level argument to specify which index level to use for membership tests
(GH7892, GH7890)
In [1]: idx = MultiIndex.from_product([[0, 1], ['a', 'b', 'c']])
In [2]: idx.values
Out[2]: array([(0, 'a'), (0, 'b'), (0, 'c'), (1, 'a'), (1, 'b'), (1, 'c')], dtype=object)
In [3]: idx.isin(['a', 'c', 'e'], level=1)
Out[3]: array([ True, False, True, True, False,

True], dtype=bool)

Index now supports duplicated and drop_duplicates. (GH4060)

In [130]: idx = Index([1, 2, 3, 4, 1, 2])
In [131]: idx
Out[131]: Int64Index([1, 2, 3, 4, 1, 2], dtype='int64')
In [132]: idx.duplicated()
Out[132]: array([False, False, False, False,

True,

True], dtype=bool)

In [133]: idx.drop_duplicates()
Out[133]: Int64Index([1, 2, 3, 4], dtype='int64')

add copy=True argument to pd.concat to enable pass thru of complete blocks (GH8252)
Added support for numpy 1.8+ data types (bool_, int_, float_, string_) for conversion to R dataframe
(GH8400)

1.7.4 Performance
Performance improvements in DatetimeIndex.__iter__ to allow faster iteration (GH7683)
Performance improvements in Period creation (and PeriodIndex setitem) (GH5155)
Improvements in Series.transform for significant performance gains (revised) (GH6496)
Performance improvements in StataReader when reading large files (GH8040, GH8073)
Performance improvements in StataWriter when writing large files (GH8079)
Performance and memory usage improvements in multi-key groupby (GH8128)
Performance improvements in groupby .agg and .apply where builtins max/min were not mapped to
numpy/cythonized versions (GH7722)
Performance improvement in writing to sql (to_sql) of up to 50% (GH8208).
Performance benchmarking of groupby for large value of ngroups (GH6787)

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Performance improvement in CustomBusinessDay, CustomBusinessMonth (GH8236)

Performance improvement for MultiIndex.values for multi-level indexes containing datetimes (GH8543)

1.7.5 Bug Fixes

Bug in pivot_table, when using margins and a dict aggfunc (GH8349)
Bug in read_csv where squeeze=True would return a view (GH8217)
Bug in checking of table name in read_sql in certain cases (GH7826).
Bug in DataFrame.groupby where Grouper does not recognize level when frequency is specified
(GH7885)
Bug in multiindexes dtypes getting mixed up when DataFrame is saved to SQL table (GH8021)
Bug in Series 0-division with a float and integer operand dtypes (GH7785)
Bug in Series.astype("unicode") not calling unicode on the values correctly (GH7758)
Bug in DataFrame.as_matrix() with mixed datetime64[ns] and timedelta64[ns] dtypes
(GH7778)
Bug in HDFStore.select_column() not preserving UTC timezone info when selecting a
DatetimeIndex (GH7777)
Bug in to_datetime when format=%Y%m%d and coerce=True are specified, where previously an
object array was returned (rather than a coerced time-series with NaT), (GH7930)
Bug in DatetimeIndex and PeriodIndex in-place addition and subtraction cause different result from
normal one (GH6527)
Bug in adding and subtracting PeriodIndex with PeriodIndex raise TypeError (GH7741)
Bug in combine_first with PeriodIndex data raises TypeError (GH3367)
Bug in multi-index slicing with missing indexers (GH7866)
Bug in multi-index slicing with various edge cases (GH8132)
Regression in multi-index indexing with a non-scalar type object (GH7914)
Bug in Timestamp comparisons with == and int64 dtype (GH8058)
Bug in pickles contains DateOffset may raise AttributeError when normalize attribute is reffered
internally (GH7748)
Bug in Panel when using major_xs and copy=False is passed (deprecation warning fails because of
missing warnings) (GH8152).
Bug in pickle deserialization that failed for pre-0.14.1 containers with dup items trying to avoid ambiguity when
matching block and manager items, when theres only one block theres no ambiguity (GH7794)
Bug in putting a PeriodIndex into a Series would convert to int64 dtype, rather than object of
Periods (GH7932)
Bug in HDFStore iteration when passing a where (GH8014)
Bug in DataFrameGroupby.transform when transforming with a passed non-sorted key (GH8046,
GH8430)
Bug in repeated timeseries line and area plot may result in ValueError or incorrect kind (GH7733)
Bug in inference in a MultiIndex with datetime.date inputs (GH7888)

1.7. v0.15.0 (October 18, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug in get where an IndexError would not cause the default value to be returned (GH7725)
Bug in offsets.apply, rollforward and rollback may reset nanosecond (GH7697)
Bug in offsets.apply, rollforward and rollback may raise AttributeError if Timestamp
has dateutil tzinfo (GH7697)
Bug in sorting a multi-index frame with a Float64Index (GH8017)
Bug in inconsistent panel setitem with a rhs of a DataFrame for alignment (GH7763)
Bug in is_superperiod and is_subperiod cannot handle higher frequencies than S (GH7760, GH7772,
GH7803)
Bug in 32-bit platforms with Series.shift (GH8129)
Bug in PeriodIndex.unique returns int64 np.ndarray (GH7540)
Bug in groupby.apply with a non-affecting mutation in the function (GH8467)
Bug in DataFrame.reset_index which has MultiIndex
DatetimeIndex with tz raises ValueError (GH7746, GH7793)

contains

PeriodIndex

Bug in DataFrame.plot with subplots=True may draw unnecessary minor xticks and yticks (GH7801)
Bug in StataReader which did not read variable labels in 117 files due to difference between Stata documentation and implementation (GH7816)
Bug in StataReader where strings were always converted to 244 characters-fixed width irrespective of underlying string size (GH7858)
Bug in DataFrame.plot and Series.plot may ignore rot and fontsize keywords (GH7844)
Bug in DatetimeIndex.value_counts doesnt preserve tz (GH7735)
Bug in PeriodIndex.value_counts results in Int64Index (GH7735)
Bug in DataFrame.join when doing left join on index and there are multiple matches (GH5391)
Bug in GroupBy.transform() where int groups with a transform that didnt preserve the index were incorrectly truncated (GH7972).
Bug in groupby where callable objects without name attributes would take the wrong path, and produce a
DataFrame instead of a Series (GH7929)
Bug in groupby error message when a DataFrame grouping column is duplicated (GH7511)
Bug in read_html where the infer_types argument forced coercion of date-likes incorrectly (GH7762,
GH7032).
Bug in Series.str.cat with an index which was filtered as to not include the first item (GH7857)
Bug in Timestamp cannot parse nanosecond from string (GH7878)
Bug in Timestamp with string offset and tz results incorrect (GH7833)
Bug in tslib.tz_convert and tslib.tz_convert_single may return different results (GH7798)
Bug in DatetimeIndex.intersection of non-overlapping timestamps with tz raises IndexError
(GH7880)
Bug in alignment with TimeOps and non-unique indexes (GH8363)
Bug in GroupBy.filter() where fast path vs. slow path made the filter return a non scalar value that
appeared valid but wasnt (GH7870).
Bug in date_range()/DatetimeIndex() when the timezone was inferred from input dates yet incorrect
times were returned when crossing DST boundaries (GH7835, GH7901).

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug in to_excel() where a negative sign was being prepended to positive infinity and was absent for negative
infinity (GH7949)
Bug in area plot draws legend with incorrect alpha when stacked=True (GH8027)
Period and PeriodIndex addition/subtraction with np.timedelta64 results in incorrect internal representations (GH7740)
Bug in Holiday with no offset or observance (GH7987)
Bug in DataFrame.to_latex formatting when columns or index is a MultiIndex (GH7982).
Bug in DateOffset around Daylight Savings Time produces unexpected results (GH5175).
Bug in DataFrame.shift where empty columns would throw ZeroDivisionError on numpy 1.7
(GH8019)
Bug in installation where html_encoding/*.html wasnt installed and therefore some tests were not running correctly (GH7927).
Bug in read_html where bytes objects were not tested for in _read (GH7927).
Bug in DataFrame.stack() when one of the column levels was a datelike (GH8039)
Bug in broadcasting numpy scalars with DataFrame (GH8116)
Bug in pivot_table performed with nameless index and columns raises KeyError (GH8103)
Bug in DataFrame.plot(kind=scatter) draws points and errorbars with different colors when the
color is specified by c keyword (GH8081)
Bug in Float64Index where iat and at were not testing and were failing (GH8092).
Bug in DataFrame.boxplot() where y-limits were not set correctly when producing multiple axes
(GH7528, GH5517).
Bug in read_csv where line comments were not handled correctly given a custom line terminator or
delim_whitespace=True (GH8122).
Bug in read_html where empty tables caused a StopIteration (GH7575)
Bug in casting when setting a column in a same-dtype block (GH7704)
Bug in accessing groups from a GroupBy when the original grouper was a tuple (GH8121).
Bug in .at that would accept integer indexers on a non-integer index and do fallback (GH7814)
Bug with kde plot and NaNs (GH8182)
Bug in GroupBy.count with float32 data type were nan values were not excluded (GH8169).
Bug with stacked barplots and NaNs (GH8175).
Bug in resample with non evenly divisible offsets (e.g. 7s) (GH8371)
Bug in interpolation methods with the limit keyword when no values needed interpolating (GH7173).
Bug where col_space was ignored in DataFrame.to_string() when header=False (GH8230).
Bug with DatetimeIndex.asof incorrectly matching partial strings and returning the wrong date
(GH8245).
Bug in plotting methods modifying the global matplotlib rcParams (GH8242).
Bug in DataFrame.__setitem__ that caused errors when setting a dataframe column to a sparse array
(GH8131)
Bug where Dataframe.boxplot() failed when entire column was empty (GH8181).

1.7. v0.15.0 (October 18, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug with messed variables in radviz visualization (GH8199).

Bug in interpolation methods with the limit keyword when no values needed interpolating (GH7173).
Bug where col_space was ignored in DataFrame.to_string() when header=False (GH8230).
Bug in to_clipboard that would clip long column data (GH8305)
Bug in DataFrame terminal display: Setting max_column/max_rows to zero did not trigger auto-resizing of
dfs to fit terminal width/height (GH7180).
Bug in OLS where running with cluster and nw_lags parameters did not work correctly, but also did not
throw an error (GH5884).
Bug in DataFrame.dropna that interpreted non-existent columns in the subset argument as the last column
(GH8303)
Bug in Index.intersection on non-monotonic non-unique indexes (GH8362).
Bug in masked series assignment where mismatching types would break alignment (GH8387)
Bug in NDFrame.equals gives false negatives with dtype=object (GH8437)
Bug in assignment with indexer where type diversity would break alignment (GH8258)
Bug in NDFrame.loc indexing when row/column names were lost when target was a list/ndarray (GH6552)
Regression in NDFrame.loc indexing when rows/columns were converted to Float64Index if target was an
empty list/ndarray (GH7774)
Bug in Series that allows it to be indexed by a DataFrame which has unexpected results. Such indexing is
no longer permitted (GH8444)
Bug in item assignment of a DataFrame with multi-index columns where right-hand-side columns were not
aligned (GH7655)
Suppress FutureWarning generated by NumPy when comparing object arrays containing NaN for equality
(GH7065)
Bug in DataFrame.eval() where the dtype of the not operator (~) was not correctly inferred as bool.

1.8 v0.14.1 (July 11, 2014)

This is a minor release from 0.14.0 and includes a small number of API changes, several new features, enhancements,
and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this
version.
Highlights include:
New methods select_dtypes() to select columns based on the dtype and sem() to calculate the
standard error of the mean.
Support for dateutil timezones (see docs).
Support for ignoring full line comments in the read_csv() text parser.
New documentation section on Options and Settings.
Lots of bug fixes.
Enhancements
API Changes
Performance Improvements
96

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Experimental Changes
Bug Fixes

1.8.1 API changes

Openpyxl now raises a ValueError on construction of the openpyxl writer instead of warning on pandas import
(GH7284).
For StringMethods.extract, when no match is found, the result - only containing NaN values - now also
has dtype=object instead of float (GH7242)
Period objects no longer raise a TypeError when compared using == with another object that isnt a
Period. Instead when comparing a Period with another object using == if the other object isnt a Period
False is returned. (GH7376)
Previously, the behaviour on resetting the time or not in offsets.apply, rollforward and rollback
operations differed between offsets. With the support of the normalize keyword for all offsets(see below) with a default value of False (preserve time), the behaviour changed for certain offsets (BusinessMonthBegin, MonthEnd, BusinessMonthEnd, CustomBusinessMonthEnd, BusinessYearBegin, LastWeekOfMonth,
FY5253Quarter, LastWeekOfMonth, Easter):
In [6]: from pandas.tseries import offsets
In [7]: d = pd.Timestamp('2014-01-01 09:00')
# old behaviour < 0.14.1
In [8]: d + offsets.MonthEnd()
Out[8]: Timestamp('2014-01-31 00:00:00')

Starting from 0.14.1 all offsets preserve time by default.

normalize=True

The old behaviour can be obtained with

# new behaviour
In [1]: d + offsets.MonthEnd()
Out[1]: Timestamp('2014-01-31 09:00:00')
In [2]: d + offsets.MonthEnd(normalize=True)
Out[2]: Timestamp('2014-01-31 00:00:00')

Note that for the other offsets the default behaviour did not change.
Add back #N/A N/A as a default NA value in text parsing, (regresion from 0.12) (GH5521)
Raise a TypeError on inplace-setting with a .where and a non np.nan value as this is inconsistent with a
set-item expression like df[mask] = None (GH7656)

1.8.2 Enhancements
Add dropna argument to value_counts and nunique (GH5569).
Add select_dtypes() method to allow selection of columns based on dtype (GH7316). See the docs.
All offsets suppports the normalize keyword to specify whether offsets.apply, rollforward
and rollback resets the time (hour, minute, etc) or not (default False, preserves time) (GH7156):
In [3]: import pandas.tseries.offsets as offsets
In [4]: day = offsets.Day()

1.8. v0.14.1 (July 11, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [5]: day.apply(Timestamp('2014-01-01 09:00'))

Out[5]: Timestamp('2014-01-02 09:00:00')
In [6]: day = offsets.Day(normalize=True)
In [7]: day.apply(Timestamp('2014-01-01 09:00'))
Out[7]: Timestamp('2014-01-02 00:00:00')

PeriodIndex is represented as the same format as DatetimeIndex (GH7601)

StringMethods now work on empty Series (GH7242)
The file parsers read_csv and read_table now ignore line comments provided by the parameter comment,
which accepts only a single character for the C reader. In particular, they allow for comments before file data
begins (GH2685)
Add NotImplementedError for simultaneous use of chunksize and nrows for read_csv() (GH6774).
Tests for basic reading of public S3 buckets now exist (GH7281).
read_html now sports an encoding argument that is passed to the underlying parser library. You can use
this to read non-ascii encoded web pages (GH7323).
read_excel now supports reading from URLs in the same way that read_csv does. (GH6809)
Support for dateutil timezones, which can now be used in the same way as pytz timezones across pandas.
(GH4688)
In [8]: rng = date_range('3/6/2012 00:00', periods=10, freq='D',
...:
tz='dateutil/Europe/London')
...:
In [9]: rng.tz
Out[9]: tzfile('/usr/share/zoneinfo/Europe/London')

See the docs.

Implemented sem (standard error of the mean) operation for Series, DataFrame, Panel, and Groupby
(GH6897)
Add nlargest and nsmallest to the Series groupby whitelist, which means you can now use these
methods on a SeriesGroupBy object (GH7053).
All offsets apply, rollforward and rollback can now handle np.datetime64, previously results in
ApplyTypeError (GH7452)
Period and PeriodIndex can contain NaT in its values (GH7485)
Support pickling Series, DataFrame and Panel objects with non-unique labels along item axis (index,
columns and items respectively) (GH7370).
Improved inference of datetime/timedelta with mixed null objects. Regression from 0.13.1 in interpretation of
an object Index with all null elements (GH7431)

1.8.3 Performance
Improvements in dtype inference for numeric operations involving yielding performance gains for dtypes:
int64, timedelta64, datetime64 (GH7223)
Improvements in Series.transform for significant performance gains (GH6496)

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Improvements in DataFrame.transform with ufuncs and built-in grouper functions for signifcant performance
gains (GH7383)
Regression in groupby aggregation of datetime64 dtypes (GH7555)
Improvements in MultiIndex.from_product for large iterables (GH7627)

1.8.4 Experimental
pandas.io.data.Options has a new method, get_all_data method, and now consistently returns a
multi-indexed DataFrame, see the docs. (GH5602)
io.gbq.read_gbq and io.gbq.to_gbq were refactored to remove the dependency on the Google
bq.py command line client. This submodule now uses httplib2 and the Google apiclient and
oauth2client API client libraries which should be more stable and, therefore, reliable than bq.py. See the
docs. (GH6937).

1.8.5 Bug Fixes

Bug in DataFrame.where with a symmetric shaped frame and a passed other of a DataFrame (GH7506)
Bug in Panel indexing with a multi-index axis (GH7516)
Regression in datetimelike slice indexing with a duplicated index and non-exact end-points (GH7523)
Bug in setitem with list-of-lists and single vs mixed types (GH7551:)
Bug in timeops with non-aligned Series (GH7500)
Bug in timedelta inference when assigning an incomplete Series (GH7592)
Bug in groupby .nth with a Series and integer-like column name (GH7559)
Bug in Series.get with a boolean accessor (GH7407)
Bug in value_counts where NaT did not qualify as missing (NaN) (GH7423)
Bug in to_timedelta that accepted invalid units and misinterpreted m/h (GH7611, GH6423)
Bug in line plot doesnt set correct xlim if secondary_y=True (GH7459)
Bug in grouped hist and scatter plots use old figsize default (GH7394)
Bug in plotting subplots with DataFrame.plot, hist clears passed ax even if the number of subplots is
one (GH7391).
Bug in plotting subplots with DataFrame.boxplot with by kw raises ValueError if the number of
subplots exceeds 1 (GH7391).
Bug in subplots displays ticklabels and labels in different rule (GH5897)
Bug in Panel.apply with a multi-index as an axis (GH7469)
Bug in DatetimeIndex.insert doesnt preserve name and tz (GH7299)
Bug in DatetimeIndex.asobject doesnt preserve name (GH7299)
Bug in multi-index slicing with datetimelike ranges (strings and Timestamps), (GH7429)
Bug in Index.min and max doesnt handle nan and NaT properly (GH7261)
Bug in PeriodIndex.min/max results in int (GH7609)
Bug in resample where fill_method was ignored if you passed how (GH2073)

1.8. v0.14.1 (July 11, 2014)

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug in TimeGrouper doesnt exclude column specified by key (GH7227)

Bug in DataFrame and Series bar and barh plot raises TypeError when bottom and left keyword is
specified (GH7226)
Bug in DataFrame.hist raises TypeError when it contains non numeric column (GH7277)
Bug in Index.delete does not preserve name and freq attributes (GH7302)
Bug in DataFrame.query()/eval where local string variables with the @ sign were being treated as
temporaries attempting to be deleted (GH7300).
Bug in Float64Index which didnt allow duplicates (GH7149).
Bug in DataFrame.replace() where truthy values were being replaced (GH7140).
Bug in StringMethods.extract() where a single match group Series would use the matchers name
instead of the group name (GH7313).
Bug in isnull() when mode.use_inf_as_null == True where isnull wouldnt test True when it
encountered an inf/-inf (GH7315).
Bug in inferred_freq results in None for eastern hemisphere timezones (GH7310)
Bug in Easter returns incorrect date when offset is negative (GH7195)
Bug in broadcasting with .div, integer dtypes and divide-by-zero (GH7325)
Bug in CustomBusinessDay.apply raiases NameError when np.datetime64 object is passed
(GH7196)
Bug in MultiIndex.append, concat and pivot_table dont preserve timezone (GH6606)
Bug in .loc with a list of indexers on a single-multi index level (that is not nested) (GH7349)
Bug in Series.map when mapping a dict with tuple keys of different lengths (GH7333)
Bug all StringMethods now work on empty Series (GH7242)
Fix delegation of read_sql to read_sql_query when query does not contain select (GH7324).
Bug where a string column name assignment to a DataFrame with a Float64Index raised a TypeError
during a call to np.isnan (GH7366).
Bug where NDFrame.replace() didnt correctly replace objects with Period values (GH7379).
Bug in .ix getitem should always return a Series (GH7150)
Bug in multi-index slicing with incomplete indexers (GH7399)
Bug in multi-index slicing with a step in a sliced level (GH7400)
Bug where negative indexers in DatetimeIndex were not correctly sliced (GH7408)
Bug where NaT wasnt reprd correctly in a MultiIndex (GH7406, GH7409).
Bug where bool objects were converted to nan in convert_objects (GH7416).
Bug in quantile ignoring the axis keyword argument (:issue7306)
Bug where nanops._maybe_null_out doesnt work with complex numbers (GH7353)
Bug in several nanops functions when axis==0 for 1-dimensional nan arrays (GH7354)
Bug where nanops.nanmedian doesnt work when axis==None (GH7352)
Bug where nanops._has_infs doesnt work with many dtypes (GH7357)
Bug in StataReader.data where reading a 0-observation dta failed (GH7369)

100

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug in StataReader when reading Stata 13 (117) files containing fixed width strings (GH7360)
Bug in StataWriter where encoding was ignored (GH7286)
Bug in DatetimeIndex comparison doesnt handle NaT properly (GH7529)
Bug in passing input with tzinfo to some offsets apply, rollforward or rollback resets tzinfo or
raises ValueError (GH7465)
Bug in DatetimeIndex.to_period, PeriodIndex.asobject, PeriodIndex.to_timestamp
doesnt preserve name (GH7485)
Bug in DatetimeIndex.to_period and PeriodIndex.to_timestanp handle NaT incorrectly
(GH7228)
Bug in offsets.apply, rollforward and rollback may return normal datetime (GH7502)
Bug in resample raises ValueError when target contains NaT (GH7227)
Bug in Timestamp.tz_localize resets nanosecond info (GH7534)
Bug in DatetimeIndex.asobject raises ValueError when it contains NaT (GH7539)
Bug in Timestamp.__new__ doesnt preserve nanosecond properly (GH7610)
Bug in Index.astype(float) where it would return an object dtype Index (GH7464).
Bug in DataFrame.reset_index loses tz (GH3950)
Bug in DatetimeIndex.freqstr raises AttributeError when freq is None (GH7606)
Bug in GroupBy.size created by TimeGrouper raises AttributeError (GH7453)
Bug in single column bar plot is misaligned (GH7498).
Bug in area plot with tz-aware time series raises ValueError (GH7471)
Bug in non-monotonic Index.union may preserve name incorrectly (GH7458)
Bug in DatetimeIndex.intersection doesnt preserve timezone (GH4690)
Bug in rolling_var where a window larger than the array would raise an error(GH7297)
Bug with last plotted timeseries dictating xlim (GH2960)
Bug with secondary_y axis not being considered for timeseries xlim (GH3490)
Bug in Float64Index assignment with a non scalar indexer (GH7586)
Bug in pandas.core.strings.str_contains does not properly match in a case insensitive fashion
when regex=False and case=False (GH7505)
Bug in expanding_cov, expanding_corr, rolling_cov, and rolling_corr for two arguments
with mismatched index (GH7512)
Bug in to_sql taking the boolean column as text column (GH7678)
Bug in grouped hist doesnt handle rot kw and sharex kw properly (GH7234)
Bug in .loc performing fallback integer indexing with object dtype indices (GH7496)
Bug (regression) in PeriodIndex constructor when passed Series objects (GH7701).

1.8. v0.14.1 (July 11, 2014)

101

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.9 v0.14.0 (May 31 , 2014)

This is a major release from 0.13.1 and includes a small number of API changes, several new features, enhancements,
and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this
version.
Highlights include:
Officially support Python 3.4
SQL interfaces updated to use sqlalchemy, See Here.
Display interface changes, See Here
MultiIndexing Using Slicers, See Here.
Ability to join a singly-indexed DataFrame with a multi-indexed DataFrame, see Here
More consistency in groupby results and more flexible groupby specifications, See Here
Holiday calendars are now supported in CustomBusinessDay, see Here
Several improvements in plotting functions, including: hexbin, area and pie plots, see Here.
Performance doc section on I/O operations, See Here
Other Enhancements
API Changes
Text Parsing API Changes
Groupby API Changes
Performance Improvements
Prior Deprecations
Deprecations
Known Issues
Bug Fixes
Warning: In 0.14.0 all NDFrame based containers have undergone significant internal refactoring. Before that
each block of homogeneous data had its own labels and extra care was necessary to keep those in sync with the
parent containers labels. This should not have any visible user/API behavior changes (GH6745)

1.9.1 API changes

read_excel uses 0 as the default sheet (GH6573)
iloc will now accept out-of-bounds indexers for slices, e.g. a value that exceeds the length of the object
being indexed. These will be excluded. This will make pandas conform more with python/numpy indexing of
out-of-bounds values. A single indexer that is out-of-bounds and drops the dimensions of the object will still
raise IndexError (GH6296, GH6299). This could result in an empty axis (e.g. an empty DataFrame being
returned)
In [1]: dfl = DataFrame(np.random.randn(5,2),columns=list('AB'))
In [2]: dfl
Out[2]:

102

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

A
B
0 1.583584 -0.438313
1 -0.402537 -0.780572
2 -0.141685 0.542241
3 0.370966 -0.251642
4 0.787484 1.666563
In [3]: dfl.iloc[:,2:3]
Out[3]:
Empty DataFrame
Columns: []
Index: [0, 1, 2, 3, 4]
In [4]: dfl.iloc[:,1:3]
Out[4]:
B
0 -0.438313
1 -0.780572
2 0.542241
3 -0.251642
4 1.666563
In [5]: dfl.iloc[4:6]
Out[5]:
A
B
4 0.787484 1.666563

These are out-of-bounds selections

dfl.iloc[[4,5,6]]
IndexError: positional indexers are out-of-bounds
dfl.iloc[:,4]
IndexError: single positional indexer is out-of-bounds

Slicing with negative start, stop & step values handles corner cases better (GH6531):
df.iloc[:-len(df)] is now empty
df.iloc[len(df)::-1] now enumerates all elements in reverse
The DataFrame.interpolate() keyword downcast default has been changed from infer to None.
This is to preseve the original dtype unless explicitly requested otherwise (GH6290).
When converting a dataframe to HTML it used to return Empty DataFrame. This special case has been removed,
instead a header with the column names is returned (GH6062).
Series
and
Index
now
internall
share
more
common
operations,
e.g.
factorize(),nunique(),value_counts() are now supported on Index types as well.
The Series.weekday property from is removed from Series for API consistency.
Using a
DatetimeIndex/PeriodIndex method on a Series will now raise a TypeError. (GH4551, GH4056,
GH5519, GH6380, GH7206).
Add
is_month_start,
is_month_end,
is_quarter_start,
is_quarter_end,
is_year_start, is_year_end accessors for DateTimeIndex / Timestamp which return a
boolean array of whether the timestamp(s) are at the start/end of the month/quarter/year defined by the
frequency of the DateTimeIndex / Timestamp (GH4565, GH6998)
Local variable usage has changed in pandas.eval()/DataFrame.eval()/DataFrame.query()
(GH5987). For the DataFrame methods, two things have changed

1.9. v0.14.0 (May 31 , 2014)

103

pandas: powerful Python data analysis toolkit, Release 0.17.0

Column names are now given precedence over locals

Local variables must be referred to explicitly. This means that even if you have a local variable that is not
a column you must still refer to it with the @ prefix.
You can have an expression like df.query(@a < a) with no complaints from pandas about ambiguity of the name a.
The top-level pandas.eval() function does not allow you use the @ prefix and provides you with
an error message telling you so.
NameResolutionError was removed because it isnt necessary anymore.
Define and document the order of column vs index names in query/eval (GH6676)
concat will now concatenate mixed Series and DataFrames using the Series name or numbering columns as
needed (GH2385). See the docs
Slicing and advanced/boolean indexing operations on Index classes as well as Index.delete() and
Index.drop() methods will no longer change the type of the resulting index (GH6440, GH7040)
In [6]: i = pd.Index([1, 2, 3, 'a' , 'b', 'c'])
In [7]: i[[0,1,2]]
Out[7]: Index([1, 2, 3], dtype='object')
In [8]: i.drop(['a', 'b', 'c'])
Out[8]: Index([1, 2, 3], dtype='object')

Previously, the above operation would return Int64Index.

Index.astype()

If youd like to do this manually, use

In [9]: i[[0,1,2]].astype(np.int_)
Out[9]: Int64Index([1, 2, 3], dtype='int32')

set_index no longer converts MultiIndexes to an Index of tuples. For example, the old behavior returned an
Index in this case (GH6459):
# Old behavior, casted MultiIndex to an Index
In [10]: tuple_ind
Out[10]: Index([(u'a', u'c'), (u'a', u'd'), (u'b', u'c'), (u'b', u'd')], dtype='object')
In [11]: df_multi.set_index(tuple_ind)
Out[11]:
0
1
(a, c) 0.471435 -1.190976
(a, d) 1.432707 -0.312652
(b, c) -0.720589 0.887163
(b, d) 0.859588 -0.636524
# New behavior
In [12]: mi
Out[12]:
MultiIndex(levels=[[u'a', u'b'], [u'c', u'd']],
labels=[[0, 0, 1, 1], [0, 1, 0, 1]])
In [13]: df_multi.set_index(mi)
Out[13]:
0
1
a c 0.471435 -1.190976
d 1.432707 -0.312652

104

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

b c -0.720589 0.887163
d 0.859588 -0.636524

This also applies when passing multiple indices to set_index:

# Old output, 2-level MultiIndex of tuples
In [14]: df_multi.set_index([df_multi.index, df_multi.index])
Out[14]:
0
1
(a, c) (a, c) 0.471435 -1.190976
(a, d) (a, d) 1.432707 -0.312652
(b, c) (b, c) -0.720589 0.887163
(b, d) (b, d) 0.859588 -0.636524
# New output, 4-level MultiIndex
In [15]: df_multi.set_index([df_multi.index, df_multi.index])
Out[15]:
0
1
a c a c 0.471435 -1.190976
d a d 1.432707 -0.312652
b c b c -0.720589 0.887163
d b d 0.859588 -0.636524

pairwise keyword was added to the statistical moment functions rolling_cov, rolling_corr,
ewmcov, ewmcorr, expanding_cov, expanding_corr to allow the calculation of moving window
covariance and correlation matrices (GH4950). See Computing rolling pairwise covariances and correlations
in the docs.
In [16]: df = DataFrame(np.random.randn(10,4),columns=list('ABCD'))
In [17]: covs = rolling_cov(df[['A','B','C']], df[['B','C','D']], 5, pairwise=True)
In [18]: covs[df.index[-1]]
Out[18]:
B
C
D
A 0.128104 0.183628 -0.047358
B 0.856265 0.058945 0.145447
C 0.058945 0.335350 0.390637

Series.iteritems() is now lazy (returns an iterator rather than a list). This was the documented behavior
prior to 0.14. (GH6760)
Added nunique and value_counts functions to Index for counting unique elements. (GH6734)
stack and unstack now raise a ValueError when the level keyword refers to a non-unique item in the
Index (previously raised a KeyError). (GH6738)
drop unused order argument from Series.sort; args now are in the same order as Series.order; add
na_position arg to conform to Series.order (GH6847)
default sorting algorithm for Series.order is now quicksort, to conform with Series.sort (and
numpy defaults)
add inplace keyword to Series.order/sort to make them inverses (GH6859)
DataFrame.sort now places NaNs at the beginning or end of the sort according to the na_position
parameter. (GH3917)
accept TextFileReader in concat, which was affecting a common user idiom (GH6583), this was a
regression from 0.13.1

1.9. v0.14.0 (May 31 , 2014)

105

pandas: powerful Python data analysis toolkit, Release 0.17.0

Added factorize functions to Index and Series to get indexer and unique values (GH7090)
describe on a DataFrame with a mix of Timestamp and string like objects returns a different Index (GH7088).
Previously the index was unintentionally sorted.
Arithmetic operations with only bool dtypes now give a warning indicating that they are evaluated in Python
space for +, -, and * operations and raise for all others (GH7011, GH6762, GH7015, GH7210)
x
y
x
x

=
=
+
/

pd.Series(np.random.rand(10) > 0.5)

True
y # warning generated: should do x | y instead
y # this raises because it doesn't make sense

NotImplementedError: operator '/' not implemented for bool dtypes

In HDFStore, select_as_multiple will always raise a KeyError, when a key or the selector is not
found (GH6177)
df[col] = value and df.loc[:,col] = value are now completely equivalent; previously the
.loc would not necessarily coerce the dtype of the resultant series (GH6149)
dtypes and ftypes now return a series with dtype=object on empty containers (GH5740)
df.to_csv will now return a string of the CSV data if neither a target path nor a buffer is provided (GH6061)
pd.infer_freq() will now raise a TypeError if given an invalid Series/Index type (GH6407,
GH6463)
A tuple passed to DataFame.sort_index will be interpreted as the levels of the index, rather than requiring
a list of tuple (GH4370)
all offset operations now return Timestamp types (rather than datetime), Business/Week frequencies were
incorrect (GH4069)
to_excel now converts np.inf into a string representation, customizable by the inf_rep keyword argument (Excel has no native inf representation) (GH6782)
Replace pandas.compat.scipy.scoreatpercentile with numpy.percentile (GH6810)
.quantile on a datetime[ns] series now returns Timestamp instead of np.datetime64 objects
(GH6810)
change AssertionError to TypeError for invalid types passed to concat (GH6583)
Raise a TypeError when DataFrame is passed an iterator as the data argument (GH5357)

1.9.2 Display Changes

The default way of printing large DataFrames has changed. DataFrames exceeding max_rows and/or
max_columns are now displayed in a centrally truncated view, consistent with the printing of a
pandas.Series (GH5603).
In previous versions, a DataFrame was truncated once the dimension constraints were reached and an ellipse
(...) signaled that part of the data was cut off.

106

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

In the current version, large DataFrames are centrally truncated, showing a preview of head and tail in both
dimensions.

allow option truncate for display.show_dimensions to only show the dimensions if the frame is
truncated (GH6547).
The default for display.show_dimensions will now be truncate. This is consistent with how Series
display length.

In [19]: dfd = pd.DataFrame(np.arange(25).reshape(-1,5), index=[0,1,2,3,4], columns=[0,1,2,3,4])

# show dimensions since this is truncated
In [20]: with pd.option_context('display.max_rows', 2, 'display.max_columns', 2,
....:
'display.show_dimensions', 'truncate'):
....:
print(dfd)
....:
0 ...
4
0
0 ...
4

1.9. v0.14.0 (May 31 , 2014)

107

pandas: powerful Python data analysis toolkit, Release 0.17.0

..
4

.. ...
20 ...

..
24

[5 rows x 5 columns]
# will not show dimensions since it is not truncated
In [21]: with pd.option_context('display.max_rows', 10, 'display.max_columns', 40,
....:
'display.show_dimensions', 'truncate'):
....:
print(dfd)
....:
0
1
2
3
4
0
0
1
2
3
4
1
5
6
7
8
9
2 10 11 12 13 14
3 15 16 17 18 19
4 20 21 22 23 24

Regression in the display of a MultiIndexed Series with display.max_rows is less than the length of the
series (GH7101)
Fixed a bug in the HTML repr of a truncated Series or DataFrame not showing the class name with the large_repr
set to info (GH7105)
The verbose keyword in DataFrame.info(), which controls whether to shorten the info representation,
is now None by default. This will follow the global setting in display.max_info_columns. The global
setting can be overriden with verbose=True or verbose=False.
Fixed a bug with the info repr not honoring the display.max_info_columns setting (GH6939)
Offset/freq info now in Timestamp __repr__ (GH4553)

1.9.3 Text Parsing API Changes

read_csv()/read_table() will now be noiser w.r.t invalid options rather than falling back to the
PythonParser.
Raise
ValueError
when
sep
read_csv()/read_table() (GH6607)

specified

Raise
ValueError
when
engine=c
read_csv()/read_table() (GH6607)

with

specified

delim_whitespace=True
with

unsupported

options

in
in

Raise ValueError when fallback to python parser causes options to be ignored (GH6607)
Produce ParserWarning on fallback to python parser when no options are ignored (GH6607)
Translate sep=\s+ to delim_whitespace=True in read_csv()/read_table() if no other Cunsupported options specified (GH6607)

1.9.4 Groupby API Changes

More consistent behaviour for some groupby methods:
groupby head and tail now act more like filter rather than an aggregation:
In [22]: df = pd.DataFrame([[1, 2], [1, 4], [5, 6]], columns=['A', 'B'])
In [23]: g = df.groupby('A')

108

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [24]: g.head(1)
Out[24]:
A B
0 1 2
2 5 6

# filters DataFrame

In [25]: g.apply(lambda x: x.head(1))

Out[25]:
A B
A
1 0 1 2
5 2 5 6

# used to simply fall-through

groupby head and tail respect column selection:

In [26]: g[['B']].head(1)
Out[26]:
B
0 2
2 6

groupby nth now reduces by default; filtering can be achieved by passing as_index=False. With an
optional dropna argument to ignore NaN. See the docs.
Reducing
In [27]: df = DataFrame([[1, np.nan], [1, 4], [5, 6]], columns=['A', 'B'])
In [28]: g = df.groupby('A')
In [29]: g.nth(0)
Out[29]:
B
A
1 NaN
5
6
# this is equivalent to g.first()
In [30]: g.nth(0, dropna='any')
Out[30]:
B
A
1 4
5 6
# this is equivalent to g.last()
In [31]: g.nth(-1, dropna='any')
Out[31]:
B
A
1 4
5 6

Filtering
In [32]: gf = df.groupby('A',as_index=False)
In [33]: gf.nth(0)
Out[33]:
A
B

1.9. v0.14.0 (May 31 , 2014)

109

pandas: powerful Python data analysis toolkit, Release 0.17.0

0
2

1 NaN
5
6

In [34]: gf.nth(0, dropna='any')

Out[34]:
B
A
1 4
5 6

groupby will now not return the grouped column for non-cython functions (GH5610, GH5614, GH6732), as its
already the index
In [35]: df = DataFrame([[1, np.nan], [1, 4], [5, 6], [5, 8]], columns=['A', 'B'])
In [36]: g = df.groupby('A')
In [37]: g.count()
Out[37]:
B
A
1 1
5 2
In [38]: g.describe()
Out[38]:
B
A
1 count 1.000000
mean
4.000000
std
NaN
min
4.000000
25%
4.000000
50%
4.000000
75%
4.000000
...
...
5 mean
7.000000
std
1.414214
min
6.000000
25%
6.500000
50%
7.000000
75%
7.500000
max
8.000000
[16 rows x 1 columns]

passing as_index will leave the grouped column in-place (this is not change in 0.14.0)
In [39]: df = DataFrame([[1, np.nan], [1, 4], [5, 6], [5, 8]], columns=['A', 'B'])
In [40]: g = df.groupby('A',as_index=False)
In [41]: g.count()
Out[41]:
A B
0 1 1
1 5 2
In [42]: g.describe()

110

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Out[42]:
A
0 count 2
mean
1
std
0
min
1
25%
1
50%
1
75%
1
...
..
1 mean
5
std
0
min
5
25%
5
50%
5
75%
5
max
5

B
1.000000
4.000000
NaN
4.000000
4.000000
4.000000
4.000000
...
7.000000
1.414214
6.000000
6.500000
7.000000
7.500000
8.000000

[16 rows x 2 columns]

Allow specification of a more complex groupby via pd.Grouper, such as grouping by a Time and a string
field simultaneously. See the docs. (GH3794)
Better propagation/preservation of Series names when performing groupby operations:
SeriesGroupBy.agg will ensure that the name attribute of the original series is propagated to the
result (GH6265).
If the function provided to GroupBy.apply returns a named series, the name of the series will be kept as
the name of the column index of the DataFrame returned by GroupBy.apply (GH6124). This facilitates
DataFrame.stack operations where the name of the column index is used as the name of the inserted
column containing the pivoted data.

1.9.5 SQL
The SQL reading and writing functions now support more database flavors through SQLAlchemy (GH2717, GH4163,
GH5950, GH6292). All databases supported by SQLAlchemy can be used, such as PostgreSQL, MySQL, Oracle,
Microsoft SQL server (see documentation of SQLAlchemy on included dialects).
The functionality of providing DBAPI connection objects will only be supported for sqlite3 in the future. The
mysql flavor is deprecated.
The new functions read_sql_query() and read_sql_table() are introduced. The function read_sql()
is kept as a convenience wrapper around the other two and will delegate to specific function depending on the provided
input (database table name or sql query).
In practice, you have to provide a SQLAlchemy engine to the sql functions. To connect with SQLAlchemy you use
the create_engine() function to create an engine object from database URI. You only need to create the engine
once per database you are connecting to. For an in-memory sqlite database:
In [43]: from sqlalchemy import create_engine
# Create your connection.
In [44]: engine = create_engine('sqlite:///:memory:')

This engine can then be used to write or read data to/from this database:

1.9. v0.14.0 (May 31 , 2014)

111

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [45]: df = pd.DataFrame({'A': [1,2,3], 'B': ['a', 'b', 'c']})

In [46]: df.to_sql('db_table', engine, index=False)

You can read data from a database by specifying the table name:
In [47]: pd.read_sql_table('db_table', engine)
Out[47]:
A B
0 1 a
1 2 b
2 3 c

or by specifying a sql query:

In [48]: pd.read_sql_query('SELECT * FROM db_table', engine)
Out[48]:
A B
0 1 a
1 2 b
2 3 c

Some other enhancements to the sql functions include:

support for writing the index. This can be controlled with the index keyword (default is True).
specify the column label to use when writing the index with index_label.
specify string columns to parse as datetimes withh the parse_dates keyword in read_sql_query() and
read_sql_table().
Warning: Some of the existing functions or function aliases have been deprecated and will be removed in future
versions. This includes: tquery, uquery, read_frame, frame_query, write_frame.
Warning: The support for the mysql flavor when using DBAPI connection objects has been deprecated. MySQL
will be further supported with SQLAlchemy engines (GH6900).

1.9.6 MultiIndexing Using Slicers

In 0.14.0 we added a new way to slice multi-indexed objects. You can slice a multi-index by providing multiple
indexers.
You can provide any of the selectors as if you are indexing by label, see Selection by Label, including slices, lists of
labels, labels, and boolean indexers.
You can use slice(None) to select all the contents of that level. You do not need to specify all the deeper levels,
they will be implied as slice(None).
As usual, both sides of the slicers are included as this is label indexing.
See the docs See also issues (GH6134, GH4036, GH3057, GH2598, GH5641, GH7106)

112

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Warning: You should specify all axes in the .loc specifier, meaning the indexer for the index and for the
columns. Their are some ambiguous cases where the passed indexer could be mis-interpreted as indexing both
axes, rather than into say the MuliIndex for the rows.
You should do this:
df.loc[(slice('A1','A3'),.....),:]

rather than this:

df.loc[(slice('A1','A3'),.....)]

Warning: You will need to make sure that the selection axes are fully lexsorted!
In [49]: def mklbl(prefix,n):
....:
return ["%s%s" % (prefix,i)
....:

for i in range(n)]

In [50]: index = MultiIndex.from_product([mklbl('A',4),

....:
mklbl('B',2),
....:
mklbl('C',4),
....:
mklbl('D',2)])
....:
In [51]: columns = MultiIndex.from_tuples([('a','foo'),('a','bar'),
....:
('b','foo'),('b','bah')],
....:
names=['lvl0', 'lvl1'])
....:
In [52]: df = DataFrame(np.arange(len(index)*len(columns)).reshape((len(index),len(columns))),
....:
index=index,
....:
columns=columns).sortlevel().sortlevel(axis=1)
....:
In [53]: df
Out[53]:
lvl0
lvl1
A0 B0 C0 D0
D1
C1 D0
D1
C2 D0
D1
C3 D0
...
A3 B1 C0 D1
C1 D0
D1
C2 D0
D1
C3 D0
D1

a
bar
1
5
9
13
17
21
25
...
229
233
237
241
245
249
253

foo
0
4
8
12
16
20
24
...
228
232
236
240
244
248
252

b
bah
3
7
11
15
19
23
27
...
231
235
239
243
247
251
255

foo
2
6
10
14
18
22
26
...
230
234
238
242
246
250
254

[64 rows x 4 columns]

Basic multi-index slicing using slices, lists, and labels.

1.9. v0.14.0 (May 31 , 2014)

113

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [54]: df.loc[(slice('A1','A3'),slice(None), ['C1','C3']),:]

Out[54]:
lvl0
a
b
lvl1
bar foo bah foo
A1 B0 C1 D0
73
72
75
74
D1
77
76
79
78
C3 D0
89
88
91
90
D1
93
92
95
94
B1 C1 D0 105 104 107 106
D1 109 108 111 110
C3 D0 121 120 123 122
...
... ... ... ...
A3 B0 C1 D1 205 204 207 206
C3 D0 217 216 219 218
D1 221 220 223 222
B1 C1 D0 233 232 235 234
D1 237 236 239 238
C3 D0 249 248 251 250
D1 253 252 255 254
[24 rows x 4 columns]

You can use a pd.IndexSlice to shortcut the creation of these slices

In [55]: idx = pd.IndexSlice
In [56]: df.loc[idx[:,:,['C1','C3']],idx[:,'foo']]
Out[56]:
lvl0
a
b
lvl1
foo foo
A0 B0 C1 D0
8
10
D1
12
14
C3 D0
24
26
D1
28
30
B1 C1 D0
40
42
D1
44
46
C3 D0
56
58
...
... ...
A3 B0 C1 D1 204 206
C3 D0 216 218
D1 220 222
B1 C1 D0 232 234
D1 236 238
C3 D0 248 250
D1 252 254
[32 rows x 2 columns]

It is possible to perform quite complicated selections using this method on multiple axes at the same time.
In [57]: df.loc['A1',(slice(None),'foo')]
Out[57]:
lvl0
a
b
lvl1
foo foo
B0 C0 D0
64
66
D1
68
70
C1 D0
72
74
D1
76
78
C2 D0
80
82

114

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

D1
C3 D0
...
B1 C0 D1
C1 D0
D1
C2 D0
D1
C3 D0
D1

84
88
...
100
104
108
112
116
120
124

86
90
...
102
106
110
114
118
122
126

[16 rows x 2 columns]

In [58]: df.loc[idx[:,:,['C1','C3']],idx[:,'foo']]
Out[58]:
lvl0
a
b
lvl1
foo foo
A0 B0 C1 D0
8
10
D1
12
14
C3 D0
24
26
D1
28
30
B1 C1 D0
40
42
D1
44
46
C3 D0
56
58
...
... ...
A3 B0 C1 D1 204 206
C3 D0 216 218
D1 220 222
B1 C1 D0 232 234
D1 236 238
C3 D0 248 250
D1 252 254
[32 rows x 2 columns]

Using a boolean indexer you can provide selection related to the values.
In [59]: mask = df[('a','foo')]>200
In [60]: df.loc[idx[mask,:,['C1','C3']],idx[:,'foo']]
Out[60]:
lvl0
a
b
lvl1
foo foo
A3 B0 C1 D1 204 206
C3 D0 216 218
D1 220 222
B1 C1 D0 232 234
D1 236 238
C3 D0 248 250
D1 252 254

You can also specify the axis argument to .loc to interpret the passed slicers on a single axis.
In [61]: df.loc(axis=0)[:,:,['C1','C3']]
Out[61]:
lvl0
a
b
lvl1
bar foo bah foo
A0 B0 C1 D0
9
8
11
10
D1
13
12
15
14

1.9. v0.14.0 (May 31 , 2014)

115

pandas: powerful Python data analysis toolkit, Release 0.17.0

C3 D0
D1
B1 C1 D0
D1
C3 D0
...
A3 B0 C1 D1
C3 D0
D1
B1 C1 D0
D1
C3 D0
D1

25
29
41
45
57
...
205
217
221
233
237
249
253

24
28
40
44
56
...
204
216
220
232
236
248
252

27
31
43
47
59
...
207
219
223
235
239
251
255

26
30
42
46
58
...
206
218
222
234
238
250
254

[32 rows x 4 columns]

Furthermore you can set the values using these methods

In [62]: df2 = df.copy()
In [63]: df2.loc(axis=0)[:,:,['C1','C3']] = -10
In [64]: df2
Out[64]:
lvl0
a
lvl1
bar
A0 B0 C0 D0
1
D1
5
C1 D0 -10
D1 -10
C2 D0
17
D1
21
C3 D0 -10
...
...
A3 B1 C0 D1 229
C1 D0 -10
D1 -10
C2 D0 241
D1 245
C3 D0 -10
D1 -10

foo
0
4
-10
-10
16
20
-10
...
228
-10
-10
240
244
-10
-10

b
bah
3
7
-10
-10
19
23
-10
...
231
-10
-10
243
247
-10
-10

foo
2
6
-10
-10
18
22
-10
...
230
-10
-10
242
246
-10
-10

[64 rows x 4 columns]

You can use a right-hand-side of an alignable object as well.

In [65]: df2 = df.copy()
In [66]: df2.loc[idx[:,:,['C1','C3']],:] = df2*1000
In [67]: df2
Out[67]:
lvl0
lvl1
A0 B0 C0 D0
D1
C1 D0
D1
C2 D0

116

a
bar
1
5
9000
9000
17

foo
0
4
8000
8000
16

b
bah
3
7
11000
11000
19

foo
2
6
10000
10000
18

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

D1
C3 D0
...
A3 B1 C0 D1
C1 D0
D1
C2 D0
D1
C3 D0
D1

21
13000
...
229
121000
121000
241
245
125000
125000

20
12000
...
228
120000
120000
240
244
124000
124000

23
15000
...
231
123000
123000
243
247
127000
127000

22
14000
...
230
122000
122000
242
246
126000
126000

[64 rows x 4 columns]

1.9.7 Plotting
Hexagonal bin plots from DataFrame.plot with kind=hexbin (GH5478), See the docs.
DataFrame.plot and Series.plot now supports area plot with specifying kind=area (GH6656),
See the docs
Pie plots from Series.plot and DataFrame.plot with kind=pie (GH6976), See the docs.
Plotting with Error Bars is now supported in the .plot method of DataFrame and Series objects (GH3796,
GH6834), See the docs.
DataFrame.plot and Series.plot now support a table keyword for plotting matplotlib.Table,
See the docs. The table keyword can receive the following values.
False: Do nothing (default).
True: Draw a table using the DataFrame or Series called plot method. Data will be transposed to
meet matplotlibs default layout.
DataFrame or Series: Draw matplotlib.table using the passed data.
The data will be
drawn as displayed in print method (not transposed automatically).
Also, helper function
pandas.tools.plotting.table is added to create a table from DataFrame and Series, and
add it to an matplotlib.Axes.
plot(legend=reverse) will now reverse the order of legend labels for most plot kinds. (GH6014)
Line plot and area plot can be stacked by stacked=True (GH6656)
Following keywords are now acceptable for DataFrame.plot() with kind=bar and kind=barh:
width: Specify the bar width. In previous versions, static value 0.5 was passed to matplotlib and it cannot
be overwritten. (GH6604)
align: Specify the bar alignment. Default is center (different from matplotlib). In previous versions,
pandas passes align=edge to matplotlib and adjust the location to center by itself, and it results align
keyword is not applied as expected. (GH4525)
position: Specify relative alignments for bar plot layout. From 0 (left/bottom-end) to 1(right/top-end).
Default is 0.5 (center). (GH6604)
Because of the default align value changes, coordinates of bar plots are now located on integer values (0.0, 1.0,
2.0 ...). This is intended to make bar plot be located on the same coodinates as line plot. However, bar plot
may differs unexpectedly when you manually adjust the bar location or drawing area, such as using set_xlim,
set_ylim, etc. In this cases, please modify your script to meet with new coordinates.

1.9. v0.14.0 (May 31 , 2014)

117

pandas: powerful Python data analysis toolkit, Release 0.17.0

The parallel_coordinates() function now takes argument color instead of colors.

A
FutureWarning is raised to alert that the old colors argument will not be supported in a future release.
(GH6956)
The parallel_coordinates() and andrews_curves() functions now take positional argument
frame instead of data. A FutureWarning is raised if the old data argument is used by name. (GH6956)
DataFrame.boxplot() now supports layout keyword (GH6769)
DataFrame.boxplot() has a new keyword argument, return_type. It accepts dict, axes, or
both, in which case a namedtuple with the matplotlib axes and a dict of matplotlib Lines is returned.

1.9.8 Prior Version Deprecations/Changes

There are prior version deprecations that are taking effect as of 0.14.0.
Remove DateRange in favor of DatetimeIndex (GH6816)
Remove column keyword from DataFrame.sort (GH4370)
Remove precision keyword from set_eng_float_format() (GH395)
Remove force_unicode keyword from DataFrame.to_string(), DataFrame.to_latex(), and
DataFrame.to_html(); these function encode in unicode by default (GH2224, GH2225)
Remove nanRep keyword from DataFrame.to_csv() and DataFrame.to_string() (GH275)
Remove unique keyword from HDFStore.select_column() (GH3256)
Remove inferTimeRule keyword from Timestamp.offset() (GH391)
Remove name keyword from get_data_yahoo() and get_data_google() ( commit b921d1a )
Remove offset keyword from DatetimeIndex constructor ( commit 3136390 )
Remove time_rule from several rolling-moment statistical functions, such as rolling_sum() (GH1042)
Removed neg - boolean operations on numpy arrays in favor of inv ~, as this is going to be deprecated in numpy
1.9 (GH6960)

1.9.9 Deprecations
The pivot_table()/DataFrame.pivot_table() and crosstab() functions now take arguments
index and columns instead of rows and cols. A FutureWarning is raised to alert that the old rows
and cols arguments will not be supported in a future release (GH5505)
The DataFrame.drop_duplicates() and DataFrame.duplicated() methods now take argument
subset instead of cols to better align with DataFrame.dropna(). A FutureWarning is raised to
alert that the old cols arguments will not be supported in a future release (GH6680)
The DataFrame.to_csv() and DataFrame.to_excel() functions now takes argument columns instead of cols. A FutureWarning is raised to alert that the old cols arguments will not be supported in a
future release (GH6645)
Indexers will warn FutureWarning when used with a scalar indexer and a non-floating point Index (GH4892,
GH6960)

# non-floating point indexes can only be indexed by integers / labels

In [1]: Series(1,np.arange(5))[3.0]
pandas/core/index.py:469: FutureWarning: scalar indexers for index type Int64Index shoul
Out[1]: 1

118

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [2]: Series(1,np.arange(5)).iloc[3.0]
pandas/core/index.py:469: FutureWarning: scalar indexers for index type Int64Index shoul
Out[2]: 1

In [3]: Series(1,np.arange(5)).iloc[3.0:4]
pandas/core/index.py:527: FutureWarning: slice indexers when using iloc should be intege
Out[3]:
3
1
dtype: int64
# these are Float64Indexes, so integer or floating point is acceptable
In [4]: Series(1,np.arange(5.))[3]
Out[4]: 1
In [5]: Series(1,np.arange(5.))[3.0]
Out[6]: 1

Numpy 1.9 compat w.r.t. deprecation warnings (GH6960)

Panel.shift() now has a function signature that matches DataFrame.shift(). The old positional argument lags has been changed to a keyword argument periods with a default value of 1. A
FutureWarning is raised if the old argument lags is used by name. (GH6910)
The order keyword argument of factorize() will be removed. (GH6926).
Remove the copy keyword from DataFrame.xs(), Panel.major_xs(), Panel.minor_xs(). A
view will be returned if possible, otherwise a copy will be made. Previously the user could think that
copy=False would ALWAYS return a view. (GH6894)
The parallel_coordinates() function now takes argument color instead of colors.
A
FutureWarning is raised to alert that the old colors argument will not be supported in a future release.
(GH6956)
The parallel_coordinates() and andrews_curves() functions now take positional argument
frame instead of data. A FutureWarning is raised if the old data argument is used by name. (GH6956)
The support for the mysql flavor when using DBAPI connection objects has been deprecated. MySQL will be
further supported with SQLAlchemy engines (GH6900).
The following io.sql functions have been deprecated: tquery, uquery, read_frame, frame_query,
write_frame.
The percentile_width keyword argument in describe() has been deprecated. Use the percentiles keyword
instead, which takes a list of percentiles to display. The default output is unchanged.
The default return type of boxplot() will change from a dict to a matpltolib Axes in a future release. You
can use the future behavior now by passing return_type=axes to boxplot.

1.9.10 Known Issues

OpenPyXL 2.0.0 breaks backwards compatibility (GH7169)

1.9.11 Enhancements
DataFrame and Series will create a MultiIndex object if passed a tuples dict, See the docs (GH3323)

1.9. v0.14.0 (May 31 , 2014)

119

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [68]: Series({('a', 'b'): 1, ('a', 'a'): 0,

....:
('a', 'c'): 2, ('b', 'a'): 3, ('b', 'b'): 4})
....:
Out[68]:
a a
0
b
1
c
2
b a
3
b
4
dtype: int64
In [69]: DataFrame({('a',
....:
('a',
....:
('a',
....:
('b',
....:
('b',
....:
Out[69]:
a
b
a
b
c
a
b
A B
4
1
5
8 10
C
3
2
6
7 NaN
D NaN NaN NaN NaN
9

'b'):
'a'):
'c'):
'a'):
'b'):

{('A',
{('A',
{('A',
{('A',
{('A',

'B'):
'C'):
'B'):
'C'):
'D'):

1,
3,
5,
7,
9,

('A',
('A',
('A',
('A',
('A',

'C'):
'B'):
'C'):
'B'):
'B'):

2},
4},
6},
8},
10}})

Added the sym_diff method to Index (GH5543)

DataFrame.to_latex now takes a longtable keyword, which if True will return a table in a longtable
environment. (GH6617)
Add option to turn off escaping in DataFrame.to_latex (GH6472)
pd.read_clipboard will, if the keyword sep is unspecified, try to detect data copied from a spreadsheet
and parse accordingly. (GH6223)
Joining a singly-indexed DataFrame with a multi-indexed DataFrame (GH3662)
See the docs. Joining multi-index DataFrames on both the left and right is not yet supported ATM.
In [70]: household = DataFrame(dict(household_id = [1,2,3],
....:
male = [0,1,0],
....:
wealth = [196087.3,316478.7,294750]),
....:
columns = ['household_id','male','wealth']
....:
).set_index('household_id')
....:
In [71]: household
Out[71]:
male
household_id
1
0
2
1
3
0

wealth
196087.3
316478.7
294750.0

In [72]: portfolio = DataFrame(dict(household_id = [1,2,2,3,3,3,4],

....:
asset_id = ["nl0000301109","nl0000289783","gb00b03mlx29",
....:
"gb00b03mlx29","lu0197800237","nl0000289965",np.
....:
name = ["ABN Amro","Robeco","Royal Dutch Shell","Royal Dutch
....:
"AAB Eastern Europe Equity Fund","Postbank BioTech F
....:
share = [1.0,0.4,0.6,0.15,0.6,0.25,1.0]),
....:
columns = ['household_id','asset_id','name','share']

120

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

....:
....:

).set_index(['household_id','asset_id'])

In [73]: portfolio
Out[73]:
household_id asset_id
1
nl0000301109
2
nl0000289783
gb00b03mlx29
3
gb00b03mlx29
lu0197800237
nl0000289965
4
NaN

name

ABN Amro
Robeco
Royal Dutch Shell
Royal Dutch Shell
AAB Eastern Europe Equity Fund
Postbank BioTech Fonds
NaN

1.00
0.40
0.60
0.15
0.60
0.25
1.00

In [74]: household.join(portfolio, how='inner')

Out[74]:
male
wealth
name
household_id asset_id
1
nl0000301109
0 196087.3
ABN Amro
2
nl0000289783
1 316478.7
Robeco
gb00b03mlx29
1 316478.7
Royal Dutch Shell
3
gb00b03mlx29
0 294750.0
Royal Dutch Shell
lu0197800237
0 294750.0 AAB Eastern Europe Equity Fund
nl0000289965
0 294750.0
Postbank BioTech Fonds

share
household_id asset_id
1
nl0000301109
2
nl0000289783
gb00b03mlx29
3
gb00b03mlx29
lu0197800237
nl0000289965

1.00
0.40
0.60
0.15
0.60
0.25

quotechar, doublequote, and escapechar can now be specified when using DataFrame.to_csv
(GH5414, GH4528)
Partially sort by only the specified levels of a MultiIndex with the sort_remaining boolean kwarg.
(GH3984)
Added to_julian_date to TimeStamp and DatetimeIndex. The Julian Date is used primarily in
astronomy and represents the number of days from noon, January 1, 4713 BC. Because nanoseconds are used
to define the time in pandas the actual range of dates that you can use is 1678 AD to 2262 AD. (GH4041)
DataFrame.to_stata will now check data for compatibility with Stata data types and will upcast when
needed. When it is not possible to losslessly upcast, a warning is issued (GH6327)
DataFrame.to_stata and StataWriter will accept keyword arguments time_stamp and data_label
which allow the time stamp and dataset label to be set when creating a file. (GH6545)
pandas.io.gbq now handles reading unicode strings properly. (GH5940)
Holidays Calendars are now available and can be used with the CustomBusinessDay offset (GH6719)
Float64Index is now backed by a float64 dtype ndarray instead of an object dtype array (GH6471).
Implemented Panel.pct_change (GH6904)
Added how option to rolling-moment functions to dictate how to handle resampling; rolling_max() defaults to max, rolling_min() defaults to min, and all others default to mean (GH6297)

1.9. v0.14.0 (May 31 , 2014)

121

pandas: powerful Python data analysis toolkit, Release 0.17.0

CustomBuisnessMonthBegin and CustomBusinessMonthEnd are now available (GH6866)

Series.quantile() and DataFrame.quantile() now accept an array of quantiles.
describe() now accepts an array of percentiles to include in the summary statistics (GH4196)
pivot_table can now accept Grouper by index and columns keywords (GH6913)
In [75]: import datetime
In [76]: df = DataFrame({
....:
'Branch' : 'A A A A A B'.split(),
....:
'Buyer': 'Carl Mark Carl Carl Joe Joe'.split(),
....:
'Quantity': [1, 3, 5, 1, 8, 1],
....:
'Date' : [datetime.datetime(2013,11,1,13,0), datetime.datetime(2013,9,1,13,5),
....:
datetime.datetime(2013,10,1,20,0), datetime.datetime(2013,10,2,10,0),
....:
datetime.datetime(2013,11,1,20,0), datetime.datetime(2013,10,2,10,0)],
....:
'PayDay' : [datetime.datetime(2013,10,4,0,0), datetime.datetime(2013,10,15,13,5),
....:
datetime.datetime(2013,9,5,20,0), datetime.datetime(2013,11,2,10,0),
....:
datetime.datetime(2013,10,7,20,0), datetime.datetime(2013,9,5,10,0)]})
....:
In [77]: df
Out[77]:
Branch Buyer
0
A Carl 2013-11-01
1
A Mark 2013-09-01
2
A Carl 2013-10-01
3
A Carl 2013-10-02
4
A
Joe 2013-11-01
5
B
Joe 2013-10-02

Date
13:00:00
13:05:00
20:00:00
10:00:00
20:00:00
10:00:00

2013-10-04
2013-10-15
2013-09-05
2013-11-02
2013-10-07
2013-09-05

PayDay
00:00:00
13:05:00
20:00:00
10:00:00
20:00:00
10:00:00

Quantity
1
3
5
1
8
1

In [78]: pivot_table(df, index=Grouper(freq='M', key='Date'),

....:
columns=Grouper(freq='M', key='PayDay'),
....:
values='Quantity', aggfunc=np.sum)
....:
Out[78]:
PayDay
2013-09-30 2013-10-31 2013-11-30
Date
2013-09-30
NaN
3
NaN
2013-10-31
6
NaN
1
2013-11-30
NaN
9
NaN

Arrays of strings can be wrapped to a specified width (str.wrap) (GH6999)

Add nsmallest() and Series.nlargest() methods to Series, See the docs (GH3960)
PeriodIndex fully supports partial string indexing like DatetimeIndex (GH7043)
In [79]: prng = period_range('2013-01-01 09:00', periods=100, freq='H')
In [80]: ps = Series(np.random.randn(len(prng)), index=prng)
In [81]: ps
Out[81]:
2013-01-01 09:00
2013-01-01 10:00
2013-01-01 11:00
2013-01-01 12:00
2013-01-01 13:00
2013-01-01 14:00

122

0.755414
0.215269
0.841009
-1.445810
-1.401973
-0.100918

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

2013-01-01 15:00

-0.548242
...
2013-01-05 06:00
-0.379811
2013-01-05 07:00
0.702562
2013-01-05 08:00
-0.850346
2013-01-05 09:00
1.176812
2013-01-05 10:00
-0.524336
2013-01-05 11:00
0.700908
2013-01-05 12:00
0.984188
Freq: H, dtype: float64
In [82]: ps['2013-01-02']
Out[82]:
2013-01-02 00:00
-0.208499
2013-01-02 01:00
1.033801
2013-01-02 02:00
-2.400454
2013-01-02 03:00
2.030604
2013-01-02 04:00
-1.142631
2013-01-02 05:00
0.211883
2013-01-02 06:00
0.704721
...
2013-01-02 17:00
0.464392
2013-01-02 18:00
-3.563517
2013-01-02 19:00
1.321106
2013-01-02 20:00
0.152631
2013-01-02 21:00
0.164530
2013-01-02 22:00
-0.430096
2013-01-02 23:00
0.767369
Freq: H, dtype: float64

read_excel can now read milliseconds in Excel dates and times with xlrd >= 0.9.3. (GH5945)
pd.stats.moments.rolling_var now uses Welfords method for increased numerical stability
(GH6817)
pd.expanding_apply and pd.rolling_apply now take args and kwargs that are passed on to the func (GH6289)
DataFrame.rank() now has a percentage rank option (GH5971)
Series.rank() now has a percentage rank option (GH5971)
Series.rank() and DataFrame.rank() now accept method=dense for ranks without gaps
(GH6514)
Support passing encoding with xlwt (GH3710)
Refactor Block classes removing Block.items attributes to avoid duplication in item handling (GH6745,
GH6988).
Testing statements updated to use specialized asserts (GH6175)

1.9.12 Performance
Performance improvement when
DatetimeConverter (GH6636)

converting

DatetimeIndex

floating

ordinals

using

Performance improvement for DataFrame.shift (GH5609)

Performance improvement in indexing into a multi-indexed Series (GH5567)
Performance improvements in single-dtyped indexing (GH6484)
1.9. v0.14.0 (May 31 , 2014)

123

pandas: powerful Python data analysis toolkit, Release 0.17.0

Improve performance of DataFrame construction with certain offsets, by removing faulty caching (e.g. MonthEnd,BusinessMonthEnd), (GH6479)
Improve performance of CustomBusinessDay (GH6584)
improve performance of slice indexing on Series with string keys (GH6341, GH6372)
Performance improvement for DataFrame.from_records when reading a specified number of rows from
an iterable (GH6700)
Performance improvements in timedelta conversions for integer dtypes (GH6754)
Improved performance of compatible pickles (GH6899)
Improve performance in certain reindexing operations by optimizing take_2d (GH6749)
GroupBy.count() is now implemented in Cython and is much faster for large numbers of groups (GH7016).

1.9.13 Experimental
There are no experimental changes in 0.14.0

1.9.14 Bug Fixes

Bug in Series ValueError when index doesnt match data (GH6532)
Prevent segfault due to MultiIndex not being supported in HDFStore table format (GH1848)
Bug in pd.DataFrame.sort_index where mergesort wasnt stable when ascending=False
(GH6399)
Bug in pd.tseries.frequencies.to_offset when argument has leading zeroes (GH6391)
Bug in version string gen. for dev versions with shallow clones / install from tarball (GH6127)
Inconsistent tz parsing Timestamp / to_datetime for current year (GH5958)
Indexing bugs with reordered indexes (GH6252, GH6254)
Bug in .xs with a Series multiindex (GH6258, GH5684)
Bug in conversion of a string types to a DatetimeIndex with a specified frequency (GH6273, GH6274)
Bug in eval where type-promotion failed for large expressions (GH6205)
Bug in interpolate with inplace=True (GH6281)
HDFStore.remove now handles start and stop (GH6177)
HDFStore.select_as_multiple handles start and stop the same way as select (GH6177)
HDFStore.select_as_coordinates and select_column works with a where clause that results in
filters (GH6177)
Regression in join of non_unique_indexes (GH6329)
Issue with groupby agg with a single function and a a mixed-type frame (GH6337)
Bug in DataFrame.replace() when passing a non- bool to_replace argument (GH6332)
Raise when trying to align on different levels of a multi-index assignment (GH3738)
Bug in setting complex dtypes via boolean indexing (GH6345)

124

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug in TimeGrouper/resample when presented with a non-monotonic DatetimeIndex that would return invalid
results. (GH4161)
Bug in index name propogation in TimeGrouper/resample (GH4161)
TimeGrouper has a more compatible API to the rest of the groupers (e.g. groups was missing) (GH3881)
Bug in multiple grouping with a TimeGrouper depending on target column order (GH6764)
Bug in pd.eval when parsing strings with possible tokens like & (GH6351)
Bug correctly handle placements of -inf in Panels when dividing by integer 0 (GH6178)
DataFrame.shift with axis=1 was raising (GH6371)
Disabled clipboard tests until release time (run locally with nosetests -A disabled) (GH6048).
Bug in DataFrame.replace() when passing a nested dict that contained keys not in the values to be
replaced (GH6342)
str.match ignored the na flag (GH6609).
Bug in take with duplicate columns that were not consolidated (GH6240)
Bug in interpolate changing dtypes (GH6290)
Bug in Series.get, was using a buggy access method (GH6383)
Bug in hdfstore queries of the form where=[(date, >=, datetime(2013,1,1)),
(date, <=, datetime(2014,1,1))] (GH6313)
Bug in DataFrame.dropna with duplicate indices (GH6355)
Regression in chained getitem indexing with embedded list-like from 0.12 (GH6394)
Float64Index with nans not comparing correctly (GH6401)
eval/query expressions with strings containing the @ character will now work (GH6366).
Bug in Series.reindex when specifying a method with some nan values was inconsistent (noted on a
resample) (GH6418)
Bug in DataFrame.replace() where nested dicts were erroneously depending on the order of dictionary
keys and values (GH5338).
Perf issue in concatting with empty objects (GH3259)
Clarify sorting of sym_diff on Index objects with NaN values (GH6444)
Regression in MultiIndex.from_product with a DatetimeIndex as input (GH6439)
Bug in str.extract when passed a non-default index (GH6348)
Bug in str.split when passed pat=None and n=1 (GH6466)
Bug
in
io.data.DataReader
when
data_source="famafrench" (GH6460)

passed

"F-F_Momentum_Factor"

and

Bug in sum of a timedelta64[ns] series (GH6462)

Bug in resample with a timezone and certain offsets (GH6397)
Bug in iat/iloc with duplicate indices on a Series (GH6493)
Bug in read_html where nans were incorrectly being used to indicate missing values in text. Should use the
empty string for consistency with the rest of pandas (GH5129).
Bug in read_html tests where redirected invalid URLs would make one test fail (GH6445).

1.9. v0.14.0 (May 31 , 2014)

125

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug in multi-axis indexing using .loc on non-unique indices (GH6504)

Bug that caused _ref_locs corruption when slice indexing across columns axis of a DataFrame (GH6525)
Regression from 0.13 in the treatment of numpy datetime64 non-ns dtypes in Series creation (GH6529)
.names attribute of MultiIndexes passed to set_index are now preserved (GH6459).
Bug in setitem with a duplicate index and an alignable rhs (GH6541)
Bug in setitem with .loc on mixed integer Indexes (GH6546)
Bug in pd.read_stata which would use the wrong data types and missing values (GH6327)
Bug in DataFrame.to_stata that lead to data loss in certain cases, and could be exported using the wrong
data types and missing values (GH6335)
StataWriter replaces missing values in string columns by empty string (GH6802)
Inconsistent types in Timestamp addition/subtraction (GH6543)
Bug in preserving frequency across Timestamp addition/subtraction (GH4547)
Bug in empty list lookup caused IndexError exceptions (GH6536, GH6551)
Series.quantile raising on an object dtype (GH6555)
Bug in .xs with a nan in level when dropped (GH6574)
Bug in fillna with method=bfill/ffill and datetime64[ns] dtype (GH6587)
Bug in sql writing with mixed dtypes possibly leading to data loss (GH6509)
Bug in Series.pop (GH6600)
Bug in iloc indexing when positional indexer matched Int64Index of the corresponding axis and no reordering happened (GH6612)
Bug in fillna with limit and value specified
Bug in DataFrame.to_stata when columns have non-string names (GH4558)
Bug in compat with np.compress, surfaced in (GH6658)
Bug in binary operations with a rhs of a Series not aligning (GH6681)
Bug in DataFrame.to_stata which incorrectly handles nan values and ignores with_index keyword
argument (GH6685)
Bug in resample with extra bins when using an evenly divisible frequency (GH4076)
Bug in consistency of groupby aggregation when passing a custom function (GH6715)
Bug in resample when how=None resample freq is the same as the axis frequency (GH5955)
Bug in downcasting inference with empty arrays (GH6733)
Bug in obj.blocks on sparse containers dropping all but the last items of same for dtype (GH6748)
Bug in unpickling NaT (NaTType) (GH4606)
Bug in DataFrame.replace() where regex metacharacters were being treated as regexs even when
regex=False (GH6777).
Bug in timedelta ops on 32-bit platforms (GH6808)
Bug in setting a tz-aware index directly via .index (GH6785)
Bug in expressions.py where numexpr would try to evaluate arithmetic ops (GH6762).

126

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Bug in Makefile where it didnt remove Cython generated C files with make clean (GH6768)
Bug with numpy < 1.7.2 when reading long strings from HDFStore (GH6166)
Bug in DataFrame._reduce where non bool-like (0/1) integers were being coverted into bools. (GH6806)
Regression from 0.13 with fillna and a Series on datetime-like (GH6344)
Bug in adding np.timedelta64 to DatetimeIndex with timezone outputs incorrect results (GH6818)
Bug in DataFrame.replace() where changing a dtype through replacement would only replace the first
occurrence of a value (GH6689)
Better error message when passing a frequency of MS in Period construction (GH5332)
Bug in Series.__unicode__ when max_rows=None and the Series has more than 1000 rows. (GH6863)
Bug in groupby.get_group where a datetlike wasnt always accepted (GH5267)
Bug in groupBy.get_group created by TimeGrouper raises AttributeError (GH6914)
Bug in DatetimeIndex.tz_localize and DatetimeIndex.tz_convert converting NaT incorrectly (GH5546)
Bug in arithmetic operations affecting NaT (GH6873)
Bug in Series.str.extract where the resulting Series from a single group match wasnt renamed to
the group name
Bug in DataFrame.to_csv where setting index=False ignored the header kwarg (GH6186)
Bug in DataFrame.plot and Series.plot, where the legend behave inconsistently when plotting to the
same axes repeatedly (GH6678)
Internal tests for patching __finalize__ / bug in merge not finalizing (GH6923, GH6927)
accept TextFileReader in concat, which was affecting a common user idiom (GH6583)
Bug in C parser with leading whitespace (GH3374)
Bug in C parser with delim_whitespace=True and \r-delimited lines
Bug in python parser with explicit multi-index in row following column header (GH6893)
Bug in Series.rank and DataFrame.rank that caused small floats (<1e-13) to all receive the same rank
(GH6886)
Bug in DataFrame.apply with functions that used *args or **kwargs and returned an empty result
(GH6952)
Bug in sum/mean on 32-bit platforms on overflows (GH6915)
Moved Panel.shift to NDFrame.slice_shift and fixed to respect multiple dtypes. (GH6959)
Bug in enabling subplots=True in DataFrame.plot only has single column raises TypeError, and
Series.plot raises AttributeError (GH6951)
Bug in DataFrame.plot draws unnecessary axes when enabling subplots and kind=scatter
(GH6951)
Bug in read_csv from a filesystem with non-utf-8 encoding (GH6807)
Bug in iloc when setting / aligning (GH6766)
Bug causing UnicodeEncodeError when get_dummies called with unicode values and a prefix (GH6885)
Bug in timeseries-with-frequency plot cursor display (GH5453)
Bug surfaced in groupby.plot when using a Float64Index (GH7025)

1.9. v0.14.0 (May 31 , 2014)

127

pandas: powerful Python data analysis toolkit, Release 0.17.0

Stopped tests from failing if options data isnt able to be downloaded from Yahoo (GH7034)
Bug in parallel_coordinates and radviz where reordering of class column caused possible color/class
mismatch (GH6956)
Bug in radviz and andrews_curves where multiple values of color were being passed to plotting method
(GH6956)
Bug in Float64Index.isin() where containing nan s would make indices claim that they contained all
the things (GH7066).
Bug in DataFrame.boxplot where it failed to use the axis passed as the ax argument (GH3578)
Bug in the XlsxWriter and XlwtWriter implementations that resulted in datetime columns being formatted without the time (GH7075) were being passed to plotting method
read_fwf() treats None in colspec like regular python slices. It now reads from the beginning or until the
end of the line when colspec contains a None (previously raised a TypeError)
Bug in cache coherence with chained indexing and slicing; add _is_view property to NDFrame to correctly
predict views; mark is_copy on xs only if its an actual copy (and not a view) (GH7084)
Bug in DatetimeIndex creation from string ndarray with dayfirst=True (GH5917)
Bug in MultiIndex.from_arrays created from DatetimeIndex doesnt preserve freq and tz
(GH7090)
Bug in unstack raises ValueError when MultiIndex contains PeriodIndex (GH4342)
Bug in boxplot and hist draws unnecessary axes (GH6769)
Regression in groupby.nth() for out-of-bounds indexers (GH6621)
Bug in quantile with datetime values (GH6965)
Bug in Dataframe.set_index, reindex and pivot dont preserve DatetimeIndex and
PeriodIndex attributes (GH3950, GH5878, GH6631)
Bug in MultiIndex.get_level_values doesnt preserve DatetimeIndex and PeriodIndex attributes (GH7092)
Bug in Groupby doesnt preserve tz (GH3950)
Bug in PeriodIndex partial string slicing (GH6716)
Bug in the HTML repr of a truncated Series or DataFrame not showing the class name with the large_repr set
to info (GH7105)
Bug in DatetimeIndex specifying freq raises ValueError when passed value is too short (GH7098)
Fixed a bug with the info repr not honoring the display.max_info_columns setting (GH6939)
Bug PeriodIndex string slicing with out of bounds values (GH5407)
Fixed a memory error in the hashtable implementation/factorizer on resizing of large tables (GH7157)
Bug in isnull when applied to 0-dimensional object arrays (GH7176)
Bug in query/eval where global constants were not looked up correctly (GH7178)
Bug in recognizing out-of-bounds positional list indexers with iloc and a multi-axis tuple indexer (GH7189)
Bug in setitem with a single value, multi-index and integer indices (GH7190, GH7218)
Bug in expressions evaluation with reversed ops, showing in series-dataframe ops (GH7198, GH7192)
Bug in multi-axis indexing with > 2 ndim and a multi-index (GH7199)

128

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Fix a bug where invalid eval/query operations would blow the stack (GH5198)

1.10 v0.13.1 (February 3, 2014)

This is a minor release from 0.13.0 and includes a small number of API changes, several new features, enhancements,
and performance improvements along with a large number of bug fixes. We recommend that all users upgrade to this
version.
Highlights include:
Added infer_datetime_format keyword to read_csv/to_datetime to allow speedups for homogeneously formatted datetimes.
Will intelligently limit display precision for datetime/timedelta formats.
Enhanced Panel apply() method.
Suggested tutorials in new Tutorials section.
Our pandas ecosystem is growing, We now feature related projects in a new Pandas Ecosystem section.
Much work has been taking place on improving the docs, and a new Contributing section has been added.
Even though it may only be of interest to devs, we <3 our new CI status page: ScatterCI.
Warning: 0.13.1 fixes a bug that was caused by a combination of having numpy < 1.8, and doing chained
assignment on a string-like array. Please review the docs, chained indexing can have unexpected results and should
generally be avoided.
This would previously segfault:
In [1]: df = DataFrame(dict(A = np.array(['foo','bar','bah','foo','bar'])))
In [2]: df['A'].iloc[0] = np.nan
In [3]: df
Out[3]:
A
0 NaN
1 bar
2 bah
3 foo
4 bar

The recommended way to do this type of assignment is:

In [4]: df = DataFrame(dict(A = np.array(['foo','bar','bah','foo','bar'])))
In [5]: df.ix[0,'A'] = np.nan
In [6]: df
Out[6]:
A
0 NaN
1 bar
2 bah
3 foo
4 bar

1.10. v0.13.1 (February 3, 2014)

129

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.10.1 Output Formatting Enhancements

df.info() view now display dtype info per column (GH5682)
df.info() now honors the option max_info_rows, to disable null counts for large frames (GH5974)
In [7]: max_info_rows = pd.get_option('max_info_rows')
In [8]: df = DataFrame(dict(A = np.random.randn(10),
...:
B = np.random.randn(10),
...:
C = date_range('20130101',periods=10)))
...:
In [9]: df.iloc[3:6,[0,2]] = np.nan
# set to not display the null counts
In [10]: pd.set_option('max_info_rows',0)
In [11]: df.info()
<class 'pandas.core.frame.DataFrame'>
Int64Index: 10 entries, 0 to 9
Data columns (total 3 columns):
A
float64
B
float64
C
datetime64[ns]
dtypes: datetime64[ns](1), float64(2)
memory usage: 320.0 bytes
# this is the default (same as in 0.13.0)
In [12]: pd.set_option('max_info_rows',max_info_rows)
In [13]: df.info()
<class 'pandas.core.frame.DataFrame'>
Int64Index: 10 entries, 0 to 9
Data columns (total 3 columns):
A
7 non-null float64
B
10 non-null float64
C
7 non-null datetime64[ns]
dtypes: datetime64[ns](1), float64(2)
memory usage: 320.0 bytes

Add show_dimensions display option for the new DataFrame repr to control whether the dimensions print.
In [14]: df = DataFrame([[1, 2], [3, 4]])
In [15]: pd.set_option('show_dimensions', False)
In [16]: df
Out[16]:
0 1
0 1 2
1 3 4
In [17]: pd.set_option('show_dimensions', True)
In [18]: df
Out[18]:
0 1
0 1 2

130

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

[2 rows x 2 columns]

The ArrayFormatter for datetime and timedelta64 now intelligently limit precision based on the
values in the array (GH3401)
Previously output might look like:
age
today
diff
0 2001-01-01 00:00:00 2013-04-19 00:00:00 4491 days, 00:00:00
1 2004-06-01 00:00:00 2013-04-19 00:00:00 3244 days, 00:00:00

Now the output looks like:

In [19]: df = DataFrame([ Timestamp('20010101'),
....:
Timestamp('20040601') ], columns=['age'])
....:
In [20]: df['today'] = Timestamp('20130419')
In [21]: df['diff'] = df['today']-df['age']
In [22]: df
Out[22]:
age
today
diff
0 2001-01-01 2013-04-19 4491 days
1 2004-06-01 2013-04-19 3244 days
[2 rows x 3 columns]

1.10.2 API changes

Add -NaN and -nan to the default set of NA values (GH5952). See NA Values.
Added Series.str.get_dummies vectorized string method (GH6021), to extract dummy/indicator variables for separated string columns:
In [23]: s = Series(['a', 'a|b', np.nan, 'a|c'])
In [24]:
Out[24]:
a b
0 1 0
1 1 1
2 0 0
3 1 0

s.str.get_dummies(sep='|')
c
0
0
0
1

[4 rows x 3 columns]

Added the NDFrame.equals() method to compare if two NDFrames are equal have equal axes, dtypes, and
values. Added the array_equivalent function to compare if two ndarrays are equal. NaNs in identical
locations are treated as equal. (GH5283) See also the docs for a motivating example.
In [25]: df = DataFrame({'col':['foo', 0, np.nan]})
In [26]: df2 = DataFrame({'col':[np.nan, 0, 'foo']}, index=[2,1,0])

1.10. v0.13.1 (February 3, 2014)

131

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [27]: df.equals(df2)
Out[27]: False
In [28]: df.equals(df2.sort())
Out[28]: True
In [29]: import pandas.core.common as com
In [30]: com.array_equivalent(np.array([0, np.nan]), np.array([0, np.nan]))
Out[30]: True
In [31]: np.array_equal(np.array([0, np.nan]), np.array([0, np.nan]))
Out[31]: False

DataFrame.apply will use the reduce argument to determine whether a Series or a DataFrame
should be returned when the DataFrame is empty (GH6007).
Previously, calling DataFrame.apply an empty DataFrame would return either a DataFrame if there
were no columns, or the function being applied would be called with an empty Series to guess whether a
Series or DataFrame should be returned:
In [32]: def applied_func(col):
....:
print("Apply function being called with: ", col)
....:
return col.sum()
....:
In [33]: empty = DataFrame(columns=['a', 'b'])
In [34]: empty.apply(applied_func)
('Apply function being called with: ', Series([], dtype: float64))
Out[34]:
a
NaN
b
NaN
dtype: float64

Now, when apply is called on an empty DataFrame: if the reduce argument is True a Series will
returned, if it is False a DataFrame will be returned, and if it is None (the default) the function being
applied will be called with an empty series to try and guess the return type.
In [35]: empty.apply(applied_func, reduce=True)
Out[35]:
a
NaN
b
NaN
dtype: float64
In [36]: empty.apply(applied_func, reduce=False)
Out[36]:
Empty DataFrame
Columns: [a, b]
Index: []
[0 rows x 2 columns]

1.10.3 Prior Version Deprecations/Changes

There are no announced changes in 0.13 or prior that are taking effect as of 0.13.1

132

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.10.4 Deprecations
There are no deprecations of prior behavior in 0.13.1

1.10.5 Enhancements
pd.read_csv and pd.to_datetime learned a new infer_datetime_format keyword which greatly
improves parsing perf in many cases. Thanks to @lexual for suggesting and @danbirken for rapidly implementing. (GH5490, GH6021)
If parse_dates is enabled and this flag is set, pandas will attempt to infer the format of the datetime strings
in the columns, and if it can be inferred, switch to a faster method of parsing them. In some cases this can
increase the parsing speed by ~5-10x.
# Try to infer the format for the index column
df = pd.read_csv('foo.csv', index_col=0, parse_dates=True,
infer_datetime_format=True)

date_format and datetime_format keywords can now be specified when writing to excel files
(GH4133)
MultiIndex.from_product convenience function for creating a MultiIndex from the cartesian product of
a set of iterables (GH6055):
In [37]: shades = ['light', 'dark']
In [38]: colors = ['red', 'green', 'blue']
In [39]: MultiIndex.from_product([shades, colors], names=['shade', 'color'])
Out[39]:
MultiIndex(levels=[[u'dark', u'light'], [u'blue', u'green', u'red']],
labels=[[1, 1, 1, 0, 0, 0], [2, 1, 0, 2, 1, 0]],
names=[u'shade', u'color'])

Panel apply() will work on non-ufuncs. See the docs.

In [40]: import pandas.util.testing as tm
In [41]: panel = tm.makePanel(5)
In [42]: panel
Out[42]:
<class 'pandas.core.panel.Panel'>
Dimensions: 3 (items) x 5 (major_axis) x 4 (minor_axis)
Items axis: ItemA to ItemC
Major_axis axis: 2000-01-03 00:00:00 to 2000-01-07 00:00:00
Minor_axis axis: A to D
In [43]: panel['ItemA']
Out[43]:
A
B
C
D
2000-01-03 0.952478 -1.239072 -1.409432 -0.014752
2000-01-04 0.988138 0.139683 1.422986 1.272395
2000-01-05 -0.072608 -0.223019 -2.147855 -1.449567
2000-01-06 -0.550603 2.123692 -1.347533 -1.195524
2000-01-07 -0.938153 0.122273 0.363565 -0.591863
[5 rows x 4 columns]

1.10. v0.13.1 (February 3, 2014)

133

pandas: powerful Python data analysis toolkit, Release 0.17.0

Specifying an apply that operates on a Series (to return a single element)

In [44]: panel.apply(lambda x: x.dtype, axis='items')
Out[44]:
A
B
C
D
2000-01-03 float64 float64 float64 float64
2000-01-04 float64 float64 float64 float64
2000-01-05 float64 float64 float64 float64
2000-01-06 float64 float64 float64 float64
2000-01-07 float64 float64 float64 float64
[5 rows x 4 columns]

A similar reduction type operation

In [45]: panel.apply(lambda x: x.sum(), axis='major_axis')
Out[45]:
ItemA
ItemB
ItemC
A 0.379252 -3.696907 3.709335
B 0.923558 0.504242 4.656781
C -3.118269 -1.545718 3.188329
D -1.979310 -0.758060 -1.436483
[4 rows x 3 columns]

This is equivalent to
In [46]: panel.sum('major_axis')
Out[46]:
ItemA
ItemB
ItemC
A 0.379252 -3.696907 3.709335
B 0.923558 0.504242 4.656781
C -3.118269 -1.545718 3.188329
D -1.979310 -0.758060 -1.436483
[4 rows x 3 columns]

A transformation operation that returns a Panel, but is computing the z-score across the major_axis
In [47]: result = panel.apply(
....:
lambda x: (x-x.mean())/x.std(),
....:
axis='major_axis')
....:
In [48]: result
Out[48]:
<class 'pandas.core.panel.Panel'>
Dimensions: 3 (items) x 5 (major_axis) x 4 (minor_axis)
Items axis: ItemA to ItemC
Major_axis axis: 2000-01-03 00:00:00 to 2000-01-07 00:00:00
Minor_axis axis: A to D
In [49]: result['ItemA']
Out[49]:
A
B
C
D
2000-01-03 1.004994 -1.166509 -0.535027 0.350970
2000-01-04 1.045875 -0.036892 1.393532 1.536326
2000-01-05 -0.170198 -0.334055 -1.037810 -0.970374
2000-01-06 -0.718186 1.588611 -0.492880 -0.736422
2000-01-07 -1.162486 -0.051156 0.672185 -0.180500

134

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

[5 rows x 4 columns]

Panel apply() operating on cross-sectional slabs. (GH1148)

In [50]: f = lambda x: ((x.T-x.mean(1))/x.std(1)).T
In [51]: result = panel.apply(f, axis = ['items','major_axis'])
In [52]: result
Out[52]:
<class 'pandas.core.panel.Panel'>
Dimensions: 4 (items) x 5 (major_axis) x 3 (minor_axis)
Items axis: A to D
Major_axis axis: 2000-01-03 00:00:00 to 2000-01-07 00:00:00
Minor_axis axis: ItemA to ItemC
In [53]: result.loc[:,:,'ItemA']
Out[53]:
A
B
C
2000-01-03 0.116579 -0.667845 -1.151538
2000-01-04 0.650448 -1.114910 0.841527
2000-01-05 -0.987433 -0.438897 -1.154468
2000-01-06 0.494000 1.060450 -0.775993
2000-01-07 -0.363770 0.013169 0.392036

D
-0.157547
0.760706
-0.015033
-1.140165
-1.123913

[5 rows x 4 columns]

This is equivalent to the following

In [54]: result = Panel(dict([ (ax,f(panel.loc[:,:,ax]))
....:
for ax in panel.minor_axis ]))
....:
In [55]: result
Out[55]:
<class 'pandas.core.panel.Panel'>
Dimensions: 4 (items) x 5 (major_axis) x 3 (minor_axis)
Items axis: A to D
Major_axis axis: 2000-01-03 00:00:00 to 2000-01-07 00:00:00
Minor_axis axis: ItemA to ItemC
In [56]: result.loc[:,:,'ItemA']
Out[56]:
A
B
C
2000-01-03 0.116579 -0.667845 -1.151538
2000-01-04 0.650448 -1.114910 0.841527
2000-01-05 -0.987433 -0.438897 -1.154468
2000-01-06 0.494000 1.060450 -0.775993
2000-01-07 -0.363770 0.013169 0.392036

D
-0.157547
0.760706
-0.015033
-1.140165
-1.123913

[5 rows x 4 columns]

1.10.6 Performance
Performance improvements for 0.13.1
Series datetime/timedelta binary operations (GH5801)

1.10. v0.13.1 (February 3, 2014)

135

pandas: powerful Python data analysis toolkit, Release 0.17.0

DataFrame count/dropna for axis=1

Series.str.contains now has a regex=False keyword which can be faster for plain (non-regex) string patterns.
(GH5879)
Series.str.extract (GH5944)
dtypes/ftypes methods (GH5968)
indexing with object dtypes (GH5968)
DataFrame.apply (GH6013)
Regression in JSON IO (GH5765)
Index construction from Series (GH6150)

1.10.7 Experimental
There are no experimental changes in 0.13.1

1.10.8 Bug Fixes

See V0.13.1 Bug Fixes for an extensive list of bugs that have been fixed in 0.13.1.
See the full release notes or issue tracker on GitHub for a complete list of all API changes, Enhancements and Bug
Fixes.

1.11 v0.13.0 (January 3, 2014)

This is a major release from 0.12.0 and includes a number of API changes, several new features and enhancements
along with a large number of bug fixes.
Highlights include:
support for a new index type Float64Index, and other Indexing enhancements
HDFStore has a new string based syntax for query specification
support for new methods of interpolation
updated timedelta operations
a new string manipulation method extract
Nanosecond support for Offsets
isin for DataFrames
Several experimental features are added, including:
new eval/query methods for expression evaluation
support for msgpack serialization
an i/o interface to Googles BigQuery
Their are several new or updated docs sections including:
Comparison with SQL, which should be useful for those familiar with SQL but still learning pandas.
Comparison with R, idiom translations from R to pandas.

136

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Enhancing Performance, ways to enhance pandas performance with eval/query.

Warning: In 0.13.0 Series has internally been refactored to no longer sub-class ndarray but instead subclass
NDFrame, similar to the rest of the pandas containers. This should be a transparent change with only very limited
API implications. See Internal Refactoring

1.11.1 API changes

read_excel now supports an integer in its sheetname argument giving the index of the sheet to read in
(GH4301).
Text parser now treats anything that reads like inf (inf, Inf, -Inf, iNf, etc.) as infinity. (GH4220,
GH4219), affecting read_table, read_csv, etc.
pandas now is Python 2/3 compatible without the need for 2to3 thanks to @jtratner. As a result, pandas now
uses iterators more extensively. This also led to the introduction of substantive parts of the Benjamin Petersons
six library into compat. (GH4384, GH4375, GH4372)
pandas.util.compat and pandas.util.py3compat have been merged into pandas.compat.
pandas.compat now includes many functions allowing 2/3 compatibility. It contains both list and iterator versions of range, filter, map and zip, plus other necessary elements for Python 3 compatibility. lmap,
lzip, lrange and lfilter all produce lists instead of iterators, for compatibility with numpy, subscripting
and pandas constructors.(GH4384, GH4375, GH4372)
Series.get with negative indexers now returns the same as [] (GH4390)
Changes to how Index and MultiIndex handle metadata (levels, labels, and names) (GH4039):
# previously, you would have set levels or labels directly
index.levels = [[1, 2, 3, 4], [1, 2, 4, 4]]
# now, you use the set_levels or set_labels methods
index = index.set_levels([[1, 2, 3, 4], [1, 2, 4, 4]])
# similarly, for names, you can rename the object
# but setting names is not deprecated
index = index.set_names(["bob", "cranberry"])
# and all methods take an inplace kwarg - but return None
index.set_names(["bob", "cranberry"], inplace=True)

All division with NDFrame objects is now truedivision, regardless of the future import. This means that operating on pandas objects will by default use floating point division, and return a floating point dtype. You can use
// and floordiv to do integer division.
Integer division
In [3]: arr = np.array([1, 2, 3, 4])
In [4]: arr2 = np.array([5, 3, 2, 1])
In [5]: arr / arr2
Out[5]: array([0, 0, 1, 4])
In [6]: Series(arr) // Series(arr2)
Out[6]:
0
0
1
0

1.11. v0.13.0 (January 3, 2014)

137

pandas: powerful Python data analysis toolkit, Release 0.17.0

2
1
3
4
dtype: int64

True Division
In [7]: pd.Series(arr) / pd.Series(arr2) # no future import required
Out[7]:
0
0.200000
1
0.666667
2
1.500000
3
4.000000
dtype: float64

Infer and downcast dtype if downcast=infer is passed to fillna/ffill/bfill (GH4604)

__nonzero__ for all NDFrame objects, will now raise a ValueError, this reverts back to (GH1073,
GH4633) behavior. See gotchas for a more detailed discussion.
This prevents doing boolean comparison on entire pandas objects, which is inherently ambiguous. These all
will raise a ValueError.
if df:
....
df1 and df2
s1 and s2

Added the .bool() method to NDFrame objects to facilitate evaluating of single-element boolean Series:
In [1]: Series([True]).bool()
Out[1]: True
In [2]: Series([False]).bool()
Out[2]: False
In [3]: DataFrame([[True]]).bool()
Out[3]: True
In [4]: DataFrame([[False]]).bool()
Out[4]: False

All non-Index NDFrames (Series, DataFrame, Panel, Panel4D, SparsePanel, etc.), now support the
entire set of arithmetic operators and arithmetic flex methods (add, sub, mul, etc.). SparsePanel does not
support pow or mod with non-scalars. (GH3765)
Series and DataFrame now have a mode() method to calculate the statistical mode(s) by axis/Series.
(GH5367)
Chained assignment will now by default warn if the user is assigning to a copy. This can be changed with the
option mode.chained_assignment, allowed options are raise/warn/None. See the docs.
In [5]: dfc = DataFrame({'A':['aaa','bbb','ccc'],'B':[1,2,3]})
In [6]: pd.set_option('chained_assignment','warn')

The following warning / exception will show if this is attempted.

In [7]: dfc.loc[0]['A'] = 1111

138

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Traceback (most recent call last)

...
SettingWithCopyWarning:
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_index,col_indexer] = value instead

Here is the correct method of assignment.

In [8]: dfc.loc[0,'A'] = 11
In [9]:
Out[9]:
A
0
11
1 bbb
2 ccc

dfc
B
1
2
3

[3 rows x 2 columns]

Panel.reindex has the following call signature Panel.reindex(items=None, major_axis=None, minor_ax

to conform with other NDFrame objects. See Internal Refactoring for more information.

Series.argmin and Series.argmax are now aliased to Series.idxmin and Series.idxmax. These return the i
min or max element respectively. Prior to 0.13.0 these would return the position of the min / max element.
(GH6214)

1.11.2 Prior Version Deprecations/Changes

These were announced changes in 0.12 or prior that are taking effect as of 0.13.0
Remove deprecated Factor (GH3650)
Remove deprecated set_printoptions/reset_printoptions (GH3046)
Remove deprecated _verbose_info (GH3215)
Remove
deprecated
read_clipboard/to_clipboard/ExcelFile/ExcelWriter
from
pandas.io.parsers (GH3717) These are available as functions in the main pandas namespace (e.g.
pd.read_clipboard)
default for tupleize_cols is now False for both to_csv and read_csv. Fair warning in 0.12
(GH3604)
default for display.max_seq_len is now 100 rather then None. This activates truncated display (...) of long
sequences in various places. (GH3391)

1.11.3 Deprecations
Deprecated in 0.13.0
deprecated iterkv, which will be removed in a future release (this was an alias of iteritems used to bypass
2to3s changes). (GH4384, GH4375, GH4372)
deprecated the string method match, whose role is now performed more idiomatically by extract. In a
future release, the default behavior of match will change to become analogous to contains, which returns
a boolean indexer. (Their distinction is strictness: match relies on re.match while contains relies on
re.search.) In this release, the deprecated behavior is the default, but the new behavior is available through
the keyword argument as_indexer=True.

1.11. v0.13.0 (January 3, 2014)

139

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.11.4 Indexing API Changes

Prior to 0.13, it was impossible to use a label indexer (.loc/.ix) to set a value that was not contained in the index
of a particular axis. (GH2578). See the docs
In the Series case this is effectively an appending operation
In [10]: s = Series([1,2,3])
In [11]: s
Out[11]:
0
1
1
2
2
3
dtype: int64
In [12]: s[5] = 5.
In [13]: s
Out[13]:
0
1
1
2
2
3
5
5
dtype: float64
In [14]: dfi = DataFrame(np.arange(6).reshape(3,2),
....:
columns=['A','B'])
....:
In [15]: dfi
Out[15]:
A B
0 0 1
1 2 3
2 4 5
[3 rows x 2 columns]

This would previously KeyError

In [16]: dfi.loc[:,'C'] = dfi.loc[:,'A']
In [17]:
Out[17]:
A B
0 0 1
1 2 3
2 4 5

dfi
C
0
2
4

[3 rows x 3 columns]

This is like an append operation.

In [18]: dfi.loc[3] = 5
In [19]: dfi
Out[19]:
A B C
0 0 1 0

140

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

1
2
3

2
4
5

3
5
5

2
4
5

[4 rows x 3 columns]

A Panel setting operation on an arbitrary axis aligns the input to the Panel
In [20]: p = pd.Panel(np.arange(16).reshape(2,4,2),
....:
items=['Item1','Item2'],
....:
major_axis=pd.date_range('2001/1/12',periods=4),
....:
minor_axis=['A','B'],dtype='float64')
....:
In [21]: p
Out[21]:
<class 'pandas.core.panel.Panel'>
Dimensions: 2 (items) x 4 (major_axis) x 2 (minor_axis)
Items axis: Item1 to Item2
Major_axis axis: 2001-01-12 00:00:00 to 2001-01-15 00:00:00
Minor_axis axis: A to B
In [22]: p.loc[:,:,'C'] = Series([30,32],index=p.items)
In [23]: p
Out[23]:
<class 'pandas.core.panel.Panel'>
Dimensions: 2 (items) x 4 (major_axis) x 3 (minor_axis)
Items axis: Item1 to Item2
Major_axis axis: 2001-01-12 00:00:00 to 2001-01-15 00:00:00
Minor_axis axis: A to C
In [24]: p.loc[:,:,'C']
Out[24]:
Item1 Item2
2001-01-12
30
32
2001-01-13
30
32
2001-01-14
30
32
2001-01-15
30
32
[4 rows x 2 columns]

1.11.5 Float64Index API Change

Added a new index type, Float64Index. This will be automatically created when passing floating values in
index creation. This enables a pure label-based slicing paradigm that makes [],ix,loc for scalar indexing
and slicing work exactly the same. See the docs, (GH263)
Construction is by default for floating type values.
In [25]: index = Index([1.5, 2, 3, 4.5, 5])
In [26]: index
Out[26]: Float64Index([1.5, 2.0, 3.0, 4.5, 5.0], dtype='float64')
In [27]: s = Series(range(5),index=index)

1.11. v0.13.0 (January 3, 2014)

141

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [28]: s
Out[28]:
1.5
0
2.0
1
3.0
2
4.5
3
5.0
4
dtype: int64

Scalar selection for [],.ix,.loc will always be label based. An integer will match an equal float index (e.g.
3 is equivalent to 3.0)
In [29]: s[3]
Out[29]: 2
In [30]: s.ix[3]
Out[30]: 2
In [31]: s.loc[3]
Out[31]: 2

The only positional indexing is via iloc

In [32]: s.iloc[3]
Out[32]: 3

A scalar index that is not found will raise KeyError

Slicing is ALWAYS on the values of the index, for [],ix,loc and ALWAYS positional with iloc
In [33]: s[2:4]
Out[33]:
2
1
3
2
dtype: int64
In [34]: s.ix[2:4]
Out[34]:
2
1
3
2
dtype: int64
In [35]: s.loc[2:4]
Out[35]:
2
1
3
2
dtype: int64
In [36]: s.iloc[2:4]
Out[36]:
3.0
2
4.5
3
dtype: int64

In float indexes, slicing using floats are allowed

In [37]: s[2.1:4.6]
Out[37]:
3.0
2
4.5
3

142

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

dtype: int64
In [38]: s.loc[2.1:4.6]
Out[38]:
3.0
2
4.5
3
dtype: int64

Indexing on other index types are preserved (and positional fallback for [],ix), with the exception, that floating
point slicing on indexes on non Float64Index will now raise a TypeError.
In [1]: Series(range(5))[3.5]
TypeError: the label [3.5] is not a proper indexer for this index type (Int64Index)
In [1]: Series(range(5))[3.5:4.5]
TypeError: the slice start [3.5] is not a proper indexer for this index type (Int64Index)

Using a scalar float indexer will be deprecated in a future version, but is allowed for now.
In [3]: Series(range(5))[3.0]
Out[3]: 3

1.11.6 HDFStore API Changes

Query Format Changes. A much more string-like query format is now supported. See the docs.
In [39]: path = 'test.h5'
In [40]: dfq = DataFrame(randn(10,4),
....:
columns=list('ABCD'),
....:
index=date_range('20130101',periods=10))
....:
In [41]: dfq.to_hdf(path,'dfq',format='table',data_columns=True)

Use boolean expressions, with in-line function evaluation.

In [42]: read_hdf(path,'dfq',
....:
where="index>Timestamp('20130104') & columns=['A', 'B']")
....:
Out[42]:
A
B
2013-01-05 -1.392054 1.153922
2013-01-06 -0.881047 0.295080
2013-01-07 -1.407085 0.126781
2013-01-08 -0.838843 0.553921
2013-01-09 1.529401 0.205455
2013-01-10 0.299071 1.076541
[6 rows x 2 columns]

Use an inline column reference

In [43]: read_hdf(path,'dfq',
....:
where="A>0 or C>0")
....:
Out[43]:
A
B

1.11. v0.13.0 (January 3, 2014)

143

pandas: powerful Python data analysis toolkit, Release 0.17.0

2013-01-01
2013-01-03
2013-01-04
2013-01-05
2013-01-06
2013-01-07
2013-01-09
2013-01-10

1.126386
0.581073
-0.275774
-1.392054
-0.881047
-1.407085
1.529401
0.299071

0.247112
2.763844
0.500483
1.153922
0.295080
0.126781
0.205455
1.076541

0.121172 0.298984
0.399325 0.668488
0.863065 -1.051628
1.181944 0.391371
1.863801 -1.712274
0.003760 -1.268994
0.313013 0.866521
0.363177 1.893680

[8 rows x 4 columns]

the format keyword now replaces the table keyword; allowed values are fixed(f) or table(t) the
same defaults as prior < 0.13.0 remain, e.g. put implies fixed format and append implies table format.
This default format can be set as an option by setting io.hdf.default_format.
In [44]: path = 'test.h5'
In [45]: df = DataFrame(randn(10,2))
In [46]: df.to_hdf(path,'df_table',format='table')
In [47]: df.to_hdf(path,'df_table2',append=True)
In [48]: df.to_hdf(path,'df_fixed')
In [49]: with get_store(path) as store:
....:
print(store)
....:
<class 'pandas.io.pytables.HDFStore'>
File path: test.h5
/df_fixed
frame
(shape->[10,2])
/df_table
frame_table (typ->appendable,nrows->10,ncols->2,indexers->[index])
/df_table2
frame_table (typ->appendable,nrows->10,ncols->2,indexers->[index])

Significant table writing performance improvements

handle a passed Series in table format (GH4330)
can now serialize a timedelta64[ns] dtype in a table (GH3577), See the docs.
added an is_open property to indicate if the underlying file handle is_open; a closed store will now report
CLOSED when viewing the store (rather than raising an error) (GH4409)
a close of a HDFStore now will close that instance of the HDFStore but will only close the actual file if
the ref count (by PyTables) w.r.t. all of the open handles are 0. Essentially you have a local instance of
HDFStore referenced by a variable. Once you close it, it will report closed. Other references (to the same
file) will continue to operate until they themselves are closed. Performing an action on a closed file will raise
ClosedFileError
In [50]: path = 'test.h5'
In [51]: df = DataFrame(randn(10,2))
In [52]: store1 = HDFStore(path)
In [53]: store2 = HDFStore(path)
In [54]: store1.append('df',df)
In [55]: store2.append('df2',df)

144

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [56]: store1
Out[56]:
<class 'pandas.io.pytables.HDFStore'>
File path: test.h5
/df
frame_table (typ->appendable,nrows->10,ncols->2,indexers->[index])
In [57]: store2
Out[57]:
<class 'pandas.io.pytables.HDFStore'>
File path: test.h5
/df
frame_table (typ->appendable,nrows->10,ncols->2,indexers->[index])
/df2
frame_table (typ->appendable,nrows->10,ncols->2,indexers->[index])
In [58]: store1.close()
In [59]: store2
Out[59]:
<class 'pandas.io.pytables.HDFStore'>
File path: test.h5
/df
frame_table (typ->appendable,nrows->10,ncols->2,indexers->[index])
/df2
frame_table (typ->appendable,nrows->10,ncols->2,indexers->[index])
In [60]: store2.close()
In [61]: store2
Out[61]:
<class 'pandas.io.pytables.HDFStore'>
File path: test.h5
File is CLOSED

removed the _quiet attribute, replace by a DuplicateWarning if retrieving duplicate rows from a table
(GH4367)
removed the warn argument from open. Instead a PossibleDataLossError exception will be raised if
you try to use mode=w with an OPEN file handle (GH4367)
allow a passed locations array or mask as a where condition (GH4467). See the docs for an example.
add the keyword dropna=True to append to change whether ALL nan rows are not written to the store
(default is True, ALL nan rows are NOT written), also settable via the option io.hdf.dropna_table
(GH4625)
pass thru store creation arguments; can be used to support in-memory stores

1.11.7 DataFrame repr Changes

The HTML and plain text representations of DataFrame now show a truncated view of the table once it exceeds
a certain size, rather than switching to the short info view (GH4886, GH5550). This makes the representation more
consistent as small DataFrames get larger.

1.11. v0.13.0 (January 3, 2014)

145

pandas: powerful Python data analysis toolkit, Release 0.17.0

To get the info view, call DataFrame.info(). If you prefer the info view as the repr for large DataFrames, you
can set this by running set_option(display.large_repr, info).

1.11.8 Enhancements
df.to_clipboard() learned a new excel keyword that lets you paste df data directly into excel (enabled
by default). (GH5070).
read_html now raises a URLError instead of catching and raising a ValueError (GH4303, GH4305)
Added a test for read_clipboard() and to_clipboard() (GH4282)
Clipboard functionality now works with PySide (GH4282)
Added a more informative error message when plot arguments contain overlapping color and style arguments
(GH4402)
to_dict now takes records as a possible outtype. Returns an array of column-keyed dictionaries. (GH4936)
NaN handing in get_dummies (GH4446) with dummy_na
# previously, nan was erroneously counted as 2 here
# now it is not counted at all
In [62]: get_dummies([1, 2, np.nan])
Out[62]:
1 2
0 1 0
1 0 1
2 0 0
[3 rows x 2 columns]
# unless requested
In [63]: get_dummies([1, 2, np.nan], dummy_na=True)
Out[63]:
1
2
NaN
0
1
0
0
1
0
1
0
2
0
0
1
[3 rows x 3 columns]

timedelta64[ns] operations. See the docs.

Warning: Most of these operations require numpy >= 1.7

146

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Using the new top-level to_timedelta, you can convert a scalar or array from the standard timedelta format
(produced by to_csv) into a timedelta type (np.timedelta64 in nanoseconds).
In [64]: to_timedelta('1 days 06:05:01.00003')
Out[64]: Timedelta('1 days 06:05:01.000030')
In [65]: to_timedelta('15.5us')
Out[65]: Timedelta('0 days 00:00:00.000015')

In [66]: to_timedelta(['1 days 06:05:01.00003','15.5us','nan'])

Out[66]: TimedeltaIndex(['1 days 06:05:01.000030', '0 days 00:00:00.000015', NaT], dtype='timede

In [67]: to_timedelta(np.arange(5),unit='s')
Out[67]: TimedeltaIndex(['00:00:00', '00:00:01', '00:00:02', '00:00:03', '00:00:04'], dtype='tim

In [68]: to_timedelta(np.arange(5),unit='d')
Out[68]: TimedeltaIndex(['0 days', '1 days', '2 days', '3 days', '4 days'], dtype='timedelta64[n

A Series of dtype timedelta64[ns] can now be divided by another timedelta64[ns] object, or

astyped to yield a float64 dtyped Series. This is frequency conversion. See the docs for the docs.
In [69]: from datetime import timedelta
In [70]: td = Series(date_range('20130101',periods=4))-Series(date_range('20121201',periods=4))
In [71]: td[2] += np.timedelta64(timedelta(minutes=5,seconds=3))
In [72]: td[3] = np.nan
In [73]: td
Out[73]:
0
31 days 00:00:00
1
31 days 00:00:00
2
31 days 00:05:03
3
NaT
dtype: timedelta64[ns]
# to days
In [74]: td / np.timedelta64(1,'D')
Out[74]:
0
31.000000
1
31.000000
2
31.003507
3
NaN
dtype: float64
In [75]: td.astype('timedelta64[D]')
Out[75]:
0
31
1
31
2
31
3
NaN
dtype: float64
# to seconds
In [76]: td / np.timedelta64(1,'s')
Out[76]:
0
2678400
1
2678400

1.11. v0.13.0 (January 3, 2014)

147

pandas: powerful Python data analysis toolkit, Release 0.17.0

2
2678703
3
NaN
dtype: float64
In [77]: td.astype('timedelta64[s]')
Out[77]:
0
2678400
1
2678400
2
2678703
3
NaN
dtype: float64

Dividing or multiplying a timedelta64[ns] Series by an integer or integer Series

In [78]: td * -1
Out[78]:
0
-31 days +00:00:00
1
-31 days +00:00:00
2
-32 days +23:54:57
3
NaT
dtype: timedelta64[ns]
In [79]: td * Series([1,2,3,4])
Out[79]:
0
31 days 00:00:00
1
62 days 00:00:00
2
93 days 00:15:09
3
NaT
dtype: timedelta64[ns]

Absolute DateOffset objects can act equivalently to timedeltas

In [80]: from pandas import offsets
In [81]: td + offsets.Minute(5) + offsets.Milli(5)
Out[81]:
0
31 days 00:05:00.005000
1
31 days 00:05:00.005000
2
31 days 00:10:03.005000
3
NaT
dtype: timedelta64[ns]

Fillna is now supported for timedeltas

In [82]: td.fillna(0)
Out[82]:
0
31 days 00:00:00
1
31 days 00:00:00
2
31 days 00:05:03
3
0 days 00:00:00
dtype: timedelta64[ns]
In [83]: td.fillna(timedelta(days=1,seconds=5))
Out[83]:
0
31 days 00:00:00
1
31 days 00:00:00
2
31 days 00:05:03
3
1 days 00:00:05
dtype: timedelta64[ns]

148

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

You can do numeric reduction operations on timedeltas.

In [84]: td.mean()
Out[84]: Timedelta('31 days 00:01:41')
In [85]: td.quantile(.1)
Out[85]: Timedelta('31 days 00:00:00')

plot(kind=kde) now accepts the optional parameters bw_method and ind, passed to
scipy.stats.gaussian_kde() (for scipy >= 0.11.0) to set the bandwidth, and to gkde.evaluate() to specify the indices at which it is evaluated, respectively. See scipy docs. (GH4298)
DataFrame constructor now accepts a numpy masked record array (GH3478)
The new vectorized string method extract return regular expression matches more conveniently.
In [86]: Series(['a1', 'b2', 'c3']).str.extract('[ab](\d)')
Out[86]:
0
1
1
2
2
NaN
dtype: object

Elements that do not match return NaN. Extracting a regular expression with more than one group returns a
DataFrame with one column per group.
In [87]: Series(['a1', 'b2', 'c3']).str.extract('([ab])(\d)')
Out[87]:
0
1
0
a
1
1
b
2
2 NaN NaN
[3 rows x 2 columns]

Elements that do not match return a row of NaN. Thus, a Series of messy strings can be converted into a likeindexed Series or DataFrame of cleaned-up or more useful strings, without necessitating get() to access tuples
or re.match objects.
Named groups like
In [88]: Series(['a1', 'b2', 'c3']).str.extract(
....:
'(?P<letter>[ab])(?P<digit>\d)')
....:
Out[88]:
letter digit
0
a
1
1
b
2
2
NaN
NaN
[3 rows x 2 columns]

and optional groups can also be used.

In [89]: Series(['a1', 'b2', '3']).str.extract(
....:
'(?P<letter>[ab])?(?P<digit>\d)')
....:
Out[89]:
letter digit
0
a
1
1
b
2

1.11. v0.13.0 (January 3, 2014)

149

pandas: powerful Python data analysis toolkit, Release 0.17.0

NaN

[3 rows x 2 columns]

read_stata now accepts Stata 13 format (GH4291)

read_fwf now infers the column specifications from the first 100 rows of the file if the data has correctly
separated and properly aligned columns using the delimiter provided to the function (GH4488).
support for nanosecond times as an offset
Warning: These operations require numpy >= 1.7
Period conversions in the range of seconds and below were reworked and extended up to nanoseconds. Periods
in the nanosecond range are now available.
In [90]: date_range('2013-01-01', periods=5, freq='5N')
Out[90]:
DatetimeIndex(['2013-01-01', '2013-01-01', '2013-01-01', '2013-01-01',
'2013-01-01'],
dtype='datetime64[ns]', freq='5N')

or with frequency as offset

In [91]: date_range('2013-01-01', periods=5, freq=pd.offsets.Nano(5))
Out[91]:
DatetimeIndex(['2013-01-01', '2013-01-01', '2013-01-01', '2013-01-01',
'2013-01-01'],
dtype='datetime64[ns]', freq='5N')

Timestamps can be modified in the nanosecond range

In [92]: t = Timestamp('20130101 09:01:02')
In [93]: t + pd.datetools.Nano(123)
Out[93]: Timestamp('2013-01-01 09:01:02.000000123')

A new method, isin for DataFrames, which plays nicely with boolean indexing. The argument to isin, what
were comparing the DataFrame to, can be a DataFrame, Series, dict, or array of values. See the docs for more.
To get the rows where any of the conditions are met:
In [94]: dfi = DataFrame({'A': [1, 2, 3, 4], 'B': ['a', 'b', 'f', 'n']})
In [95]: dfi
Out[95]:
A B
0 1 a
1 2 b
2 3 f
3 4 n
[4 rows x 2 columns]
In [96]: other = DataFrame({'A': [1, 3, 3, 7], 'B': ['e', 'f', 'f', 'e']})
In [97]: mask = dfi.isin(other)
In [98]: mask

150

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Out[98]:
A
0
True
1 False
2
True
3 False

B
False
False
True
False

[4 rows x 2 columns]
In [99]: dfi[mask.any(1)]
Out[99]:
A B
0 1 a
2 3 f
[2 rows x 2 columns]

Series now supports a to_frame method to convert it to a single-column DataFrame (GH5164)

All R datasets listed here http://stat.ethz.ch/R-manual/R-devel/library/datasets/html/00Index.html can now be
loaded into Pandas objects
# note that pandas.rpy was deprecated in v0.16.0
import pandas.rpy.common as com
com.load_data('Titanic')

tz_localize can infer a fall daylight savings transition based on the structure of the unlocalized data
(GH4230), see the docs
DatetimeIndex is now in the API documentation, see the docs
json_normalize() is a new method to allow you to create a flat table from semi-structured JSON data. See
the docs (GH1067)
Added PySide support for the qtpandas DataFrameModel and DataFrameWidget.
Python csv parser now supports usecols (GH4335)
Frequencies gained several new offsets:
LastWeekOfMonth (GH4637)
FY5253, and FY5253Quarter (GH4511)
DataFrame has a new interpolate method, similar to Series (GH4434, GH1892)
In [100]: df = DataFrame({'A': [1, 2.1, np.nan, 4.7, 5.6, 6.8],
.....:
'B': [.25, np.nan, np.nan, 4, 12.2, 14.4]})
.....:
In [101]: df.interpolate()
Out[101]:
A
B
0 1.0
0.25
1 2.1
1.50
2 3.4
2.75
3 4.7
4.00
4 5.6 12.20
5 6.8 14.40
[6 rows x 2 columns]

1.11. v0.13.0 (January 3, 2014)

151

pandas: powerful Python data analysis toolkit, Release 0.17.0

Additionally, the method argument to interpolate has been expanded to include nearest,
zero, slinear, quadratic, cubic, barycentric, krogh,
piecewise_polynomial, pchip, polynomial, spline The new methods require scipy. Consult the Scipy reference guide and documentation for more information about when the various
methods are appropriate. See the docs.
Interpolate now also accepts a limit keyword argument. This works similar to fillnas limit:
In [102]: ser = Series([1, 3, np.nan, np.nan, np.nan, 11])
In [103]: ser.interpolate(limit=2)
Out[103]:
0
1
1
3
2
5
3
7
4
NaN
5
11
dtype: float64

Added wide_to_long panel data convenience function. See the docs.

In [104]: np.random.seed(123)
In [105]: df = pd.DataFrame({"A1970" : {0 : "a", 1 : "b",
.....:
"A1980" : {0 : "d", 1 : "e",
.....:
"B1970" : {0 : 2.5, 1 : 1.2,
.....:
"B1980" : {0 : 3.2, 1 : 1.3,
.....:
"X"
: dict(zip(range(3),
.....:
})
.....:

2 : "c"},
2 : "f"},
2 : .7},
2 : .1},
np.random.randn(3)))

In [106]: df["id"] = df.index

In [107]: df
Out[107]:
A1970 A1980
0
a
d
1
b
e
2
c
f

B1970
2.5
1.2
0.7

B1980
X
3.2 -1.085631
1.3 0.997345
0.1 0.282978

id
0
1
2

[3 rows x 6 columns]
In [108]: wide_to_long(df, ["A", "B"], i="id", j="year")
Out[108]:
X A
B
id year
0 1970 -1.085631 a 2.5
1 1970 0.997345 b 1.2
2 1970 0.282978 c 0.7
0 1980 -1.085631 d 3.2
1 1980 0.997345 e 1.3
2 1980 0.282978 f 0.1
[6 rows x 3 columns]

to_csv now takes a date_format keyword argument that specifies how output datetime objects should
be formatted. Datetimes encountered in the index, columns, and values will all have this formatting applied.
(GH4313)

152

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

DataFrame.plot will scatter plot x versus y by passing kind=scatter (GH2215)

Added support for Google Analytics v3 API segment IDs that also supports v2 IDs. (GH5271)

1.11.9 Experimental
The new eval() function implements expression evaluation using numexpr behind the scenes. This results
in large speedups for complicated expressions involving large DataFrames/Series. For example,
In [109]: nrows, ncols = 20000, 100
In [110]: df1, df2, df3, df4 = [DataFrame(randn(nrows, ncols))
.....:
for _ in range(4)]
.....:
# eval with NumExpr backend
In [111]: %timeit pd.eval('df1 + df2 + df3 + df4')
100 loops, best of 3: 14.9 ms per loop
# pure Python evaluation
In [112]: %timeit df1 + df2 + df3 + df4
10 loops, best of 3: 22.8 ms per loop

For more details, see the the docs

Similar to pandas.eval, DataFrame has a new DataFrame.eval method that evaluates an expression
in the context of the DataFrame. For example,
In [113]: df = DataFrame(randn(10, 2), columns=['a', 'b'])
In [114]: df.eval('a + b')
Out[114]:
0
-0.685204
1
1.589745
2
0.325441
3
-1.784153
4
-0.432893
5
0.171850
6
1.895919
7
3.065587
8
-0.092759
9
1.391365
dtype: float64

query() method has been added that allows you to select elements of a DataFrame using a natural query
syntax nearly identical to Python syntax. For example,
In [115]: n = 20
In [116]: df = DataFrame(np.random.randint(n, size=(n, 3)), columns=['a', 'b', 'c'])
In [117]: df.query('a < b < c')
Out[117]:
a
b
c
11 1
5
8
15 8 16 19
[2 rows x 3 columns]

1.11. v0.13.0 (January 3, 2014)

153

pandas: powerful Python data analysis toolkit, Release 0.17.0

selects all the rows of df where a < b < c evaluates to True. For more details see the the docs.
pd.read_msgpack() and pd.to_msgpack() are now a supported method of serialization of arbitrary
pandas (and python objects) in a lightweight portable binary format. See the docs
Warning: Since this is an EXPERIMENTAL LIBRARY, the storage format may not be stable until a future
release.
In [118]: df = DataFrame(np.random.rand(5,2),columns=list('AB'))
In [119]: df.to_msgpack('foo.msg')
In [120]: pd.read_msgpack('foo.msg')
Out[120]:
A
B
0 0.251082 0.017357
1 0.347915 0.929879
2 0.546233 0.203368
3 0.064942 0.031722
4 0.355309 0.524575
[5 rows x 2 columns]
In [121]: s = Series(np.random.rand(5),index=date_range('20130101',periods=5))
In [122]: pd.to_msgpack('foo.msg', df, s)
In [123]: pd.read_msgpack('foo.msg')
Out[123]:
[
A
B
0 0.251082 0.017357
1 0.347915 0.929879
2 0.546233 0.203368
3 0.064942 0.031722
4 0.355309 0.524575
[5 rows x 2 columns], 2013-01-01
2013-01-02
0.227025
2013-01-03
0.383282
2013-01-04
0.193225
2013-01-05
0.110977
Freq: D, dtype: float64]

0.022321

You can pass iterator=True to iterator over the unpacked results

In [124]: for o in pd.read_msgpack('foo.msg',iterator=True):
.....:
print o
.....:
A
B
0 0.251082 0.017357
1 0.347915 0.929879
2 0.546233 0.203368
3 0.064942 0.031722
4 0.355309 0.524575
[5 rows x 2 columns]
2013-01-01
0.022321
2013-01-02
0.227025
2013-01-03
0.383282

154

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

2013-01-04
0.193225
2013-01-05
0.110977
Freq: D, dtype: float64

pandas.io.gbq provides a simple way to extract from, and load data into, Googles BigQuery Data Sets by
way of pandas DataFrames. BigQuery is a high performance SQL-like database service, useful for performing
ad-hoc queries against extremely large datasets. See the docs
from pandas.io import gbq
#
#
#
#

A query to select the average monthly temperatures in the

in the year 2000 across the USA. The dataset,
publicata:samples.gsod, is available on all BigQuery accounts,
and is based on NOAA gsod data.

query = """SELECT station_number as STATION,

month as MONTH, AVG(mean_temp) as MEAN_TEMP
FROM publicdata:samples.gsod
WHERE YEAR = 2000
GROUP BY STATION, MONTH
ORDER BY STATION, MONTH ASC"""
# Fetch the result set for this query
# Your Google BigQuery Project ID
# To find this, see your dashboard:
# https://code.google.com/apis/console/b/0/?noredirect
projectid = xxxxxxxxx;
df = gbq.read_gbq(query, project_id = projectid)
# Use pandas to process and reshape the dataset
df2 = df.pivot(index='STATION', columns='MONTH', values='MEAN_TEMP')
df3 = pandas.concat([df2.min(), df2.mean(), df2.max()],
axis=1,keys=["Min Tem", "Mean Temp", "Max Temp"])

The resulting DataFrame is:

> df3
Min Tem
MONTH
1
2
3
4
5
6
7
8
9
10
11
12

-53.336667
-49.837500
-77.926087
-82.892858
-92.378261
-77.703334
-87.821428
-89.431999
-86.611112
-78.209677
-50.125000
-50.332258

Mean Temp
39.827892
43.685219
48.708355
55.070087
61.428117
65.858888
68.169663
68.614215
63.436935
56.880838
48.861228
42.286879

1.11. v0.13.0 (January 3, 2014)

Max Temp
89.770968
93.437932
96.099998
97.317240
102.042856
102.900000
106.510714
105.500000
107.142856
92.103333
94.996428
94.396774

155

pandas: powerful Python data analysis toolkit, Release 0.17.0

Warning:
To use this module, you will need a BigQuery account.
See
<https://cloud.google.com/products/big-query> for details.
As of 10/10/13, there is a bug in Googles API preventing result sets from being larger than 100,000 rows.
A patch is scheduled for the week of 10/14/13.

1.11.10 Internal Refactoring

In 0.13.0 there is a major refactor primarily to subclass Series from NDFrame, which is the base class currently
for DataFrame and Panel, to unify methods and behaviors. Series formerly subclassed directly from ndarray.
(GH4080, GH3862, GH816)

156

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Warning: There are two potential incompatibilities from < 0.13.0

Using certain numpy functions would previously return a Series if passed a Series as an argument. This
seems only to affect np.ones_like, np.empty_like, np.diff and np.where. These now return
ndarrays.
In [125]: s = Series([1,2,3,4])

Numpy Usage
In [126]: np.ones_like(s)
Out[126]: array([1, 1, 1, 1], dtype=int64)
In [127]: np.diff(s)
Out[127]: array([1, 1, 1], dtype=int64)
In [128]: np.where(s>1,s,np.nan)
Out[128]: array([ nan,
2.,
3.,

4.])

Pandonic Usage
In [129]: Series(1,index=s.index)
Out[129]:
0
1
1
1
2
1
3
1
dtype: int64
In [130]: s.diff()
Out[130]:
0
NaN
1
1
2
1
3
1
dtype: float64
In [131]: s.where(s>1)
Out[131]:
0
NaN
1
2
2
3
3
4
dtype: float64

Passing a Series directly to a cython function expecting an ndarray type will no long work directly, you
must pass Series.values, See Enhancing Performance
Series(0.5) would previously return the scalar 0.5, instead this will return a 1-element Series
This change breaks rpy2<=2.3.8. an Issue has been opened against rpy2 and a workaround is detailed in
GH5698. Thanks @JanSchulz.
Pickle compatibility is preserved for pickles created prior to 0.13.
pd.read_pickle, see Pickling.

These must be unpickled with

Refactor of series.py/frame.py/panel.py to move common code to generic.py

added _setup_axes to created generic NDFrame structures
moved methods
* from_axes,_wrap_array,axes,ix,loc,iloc,shape,empty,swapaxes,transpose,pop
1.11. v0.13.0 (January 3, 2014)

157

pandas: powerful Python data analysis toolkit, Release 0.17.0

* __iter__,keys,__contains__,__len__,__neg__,__invert__
* convert_objects,as_blocks,as_matrix,values
* __getstate__,__setstate__ (compat remains in frame/panel)
* __getattr__,__setattr__
* _indexed_same,reindex_like,align,where,mask
* fillna,replace (Series replace is now consistent with DataFrame)
* filter (also added axis argument to selectively filter on a different axis)
* reindex,reindex_axis,take
* truncate (moved to become part of NDFrame)
These are API changes which make Panel more consistent with DataFrame
swapaxes on a Panel with the same axes specified now return a copy
support attribute access for setting
filter supports the same API as the original DataFrame filter
Reindex called with no arguments will now return a copy of the input object
TimeSeries is now an alias for Series. the property is_time_series can be used to distinguish (if
desired)
Refactor of Sparse objects to use BlockManager
Created a new block type in internals, SparseBlock, which can hold multi-dtypes and is nonconsolidatable. SparseSeries and SparseDataFrame now inherit more methods from there hierarchy (Series/DataFrame), and no longer inherit from SparseArray (which instead is the object of
the SparseBlock)
Sparse suite now supports integration with non-sparse data. Non-float sparse data is supportable (partially
implemented)
Operations on sparse structures within DataFrames should preserve sparseness, merging type operations
will convert to dense (and back to sparse), so might be somewhat inefficient
enable setitem on SparseSeries for boolean/integer/slices
SparsePanels implementation is unchanged (e.g. not using BlockManager, needs work)
added ftypes method to Series/DataFrame, similar to dtypes, but indicates if the underlying is sparse/dense
(as well as the dtype)
All NDFrame objects can now use __finalize__() to specify various values to propagate to new objects
from an existing one (e.g. name in Series will follow more automatically now)
Internal type checking is now done via a suite of generated classes, allowing isinstance(value, klass)
without having to directly import the klass, courtesy of @jtratner
Bug in Series update where the parent frame is not updating its cache based on changes (GH4080) or types
(GH3217), fillna (GH3386)
Indexing with dtype conversions fixed (GH4463, GH4204)
Refactor Series.reindex to core/generic.py (GH4604, GH4618), allow method= in reindexing on a Series to work
Series.copy no longer accepts the order parameter and is now consistent with NDFrame copy

158

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Refactor rename methods to core/generic.py; fixes Series.rename for (GH4605), and adds rename with
the same signature for Panel
Refactor clip methods to core/generic.py (GH4798)
Refactor of _get_numeric_data/_get_bool_data to core/generic.py, allowing Series/Panel functionality
Series (for index) / Panel (for items) now allow attribute access to its elements (GH1903)
In [132]: s = Series([1,2,3],index=list('abc'))
In [133]: s.b
Out[133]: 2
In [134]: s.a = 5
In [135]: s
Out[135]:
a
5
b
2
c
3
dtype: int64

1.11.11 Bug Fixes

See V0.13.0 Bug Fixes for an extensive list of bugs that have been fixed in 0.13.0.
See the full release notes or issue tracker on GitHub for a complete list of all API changes, Enhancements and Bug
Fixes.

1.12 v0.12.0 (July 24, 2013)

This is a major release from 0.11.0 and includes several new features and enhancements along with a large number of
bug fixes.
Highlights include a consistent I/O API naming scheme, routines to read html, write multi-indexes to csv files, read
& write STATA data files, read & write JSON format files, Python 3 support for HDFStore, filtering of groupby
expressions via filter, and a revamped replace routine that accepts regular expressions.

1.12.1 API changes

The I/O API is now much more consistent with a set of top level reader functions accessed like
pd.read_csv() that generally return a pandas object.
read_csv
read_excel
read_hdf
read_sql
read_json
read_html
read_stata
1.12. v0.12.0 (July 24, 2013)

159

pandas: powerful Python data analysis toolkit, Release 0.17.0

read_clipboard
The corresponding writer functions are object methods that are accessed like df.to_csv()
to_csv
to_excel
to_hdf
to_sql
to_json
to_html
to_stata
to_clipboard
Fix modulo and integer division on Series,DataFrames to act similary to float dtypes to return np.nan
or np.inf as appropriate (GH3590). This correct a numpy bug that treats integer and float dtypes
differently.
In [1]: p = DataFrame({ 'first' : [4,5,8], 'second' : [0,0,3] })
In [2]: p % 0
Out[2]:
first second
0
NaN
NaN
1
NaN
NaN
2
NaN
NaN
[3 rows x 2 columns]
In [3]: p % p
Out[3]:
first second
0
0
NaN
1
0
NaN
2
0
0
[3 rows x 2 columns]
In [4]: p / p
Out[4]:
first second
0
1
NaN
1
1
NaN
2
1
1
[3 rows x 2 columns]
In [5]: p / 0
Out[5]:
first second
0
inf
NaN
1
inf
NaN
2
inf
inf
[3 rows x 2 columns]

160

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Add squeeze keyword to groupby to allow reduction from DataFrame -> Series if groups are unique. This
is a Regression from 0.10.1. We are reverting back to the prior behavior. This means groupby will return the
same shaped objects whether the groups are unique or not. Revert this issue (GH2893) with (GH3596).
In [6]: df2 = DataFrame([{"val1": 1, "val2" : 20}, {"val1":1, "val2": 19},
...:
{"val1":1, "val2": 27}, {"val1":1, "val2": 12}])
...:
In [7]: def func(dataf):
...:
return dataf["val2"]
...:

- dataf["val2"].mean()

# squeezing the result frame to a series (because we have unique groups)

In [8]: df2.groupby("val1", squeeze=True).apply(func)
Out[8]:
0
0.5
1
-0.5
2
7.5
3
-7.5
Name: 1, dtype: float64
# no squeezing (the default, and behavior in 0.10.1)
In [9]: df2.groupby("val1").apply(func)
Out[9]:
val2
0
1
2
3
val1
1
0.5 -0.5 7.5 -7.5
[1 rows x 4 columns]

Raise on iloc when boolean indexing with a label based indexer mask e.g. a boolean Series, even with integer
labels, will raise. Since iloc is purely positional based, the labels on the Series are not alignable (GH3631)
This case is rarely used, and there are plently of alternatives. This preserves the iloc API to be purely positional
based.
In [10]: df = DataFrame(lrange(5), list('ABCDE'), columns=['a'])
In [11]: mask = (df.a%2 == 0)
In [12]: mask
Out[12]:
A
True
B
False
C
True
D
False
E
True
Name: a, dtype: bool
# this is what you should use
In [13]: df.loc[mask]
Out[13]:
a
A 0
C 2
E 4
[3 rows x 1 columns]

1.12. v0.12.0 (July 24, 2013)

161

pandas: powerful Python data analysis toolkit, Release 0.17.0

# this will work as well

In [14]: df.iloc[mask.values]
Out[14]:
a
A 0
C 2
E 4
[3 rows x 1 columns]

df.iloc[mask] will raise a ValueError

The raise_on_error argument to plotting functions is removed. Instead, plotting functions raise a
TypeError when the dtype of the object is object to remind you to avoid object arrays whenever
possible and thus you should cast to an appropriate numeric dtype if you need to plot something.
Add colormap keyword to DataFrame plotting methods. Accepts either a matplotlib colormap object (ie,
matplotlib.cm.jet) or a string name of such an object (ie, jet). The colormap is sampled to select the color for
each column. Please see Colormaps for more information. (GH3860)
DataFrame.interpolate() is now deprecated.
Please use DataFrame.fillna() and
DataFrame.replace() instead. (GH3582, GH3675, GH3676)
the method and axis arguments of DataFrame.replace() are deprecated
DataFrame.replace s infer_types parameter is removed and now performs conversion by default.
(GH3907)
Add the keyword allow_duplicates to DataFrame.insert to allow a duplicate column to be inserted
if True, default is False (same as prior to 0.12) (GH3679)
Implement __nonzero__ for NDFrame objects (GH3691, GH3696)
IO api
added top-level function read_excel to replace the following, The original API is deprecated and will
be removed in a future version
from pandas.io.parsers import ExcelFile
xls = ExcelFile('path_to_file.xls')
xls.parse('Sheet1', index_col=None, na_values=['NA'])

With
import pandas as pd
pd.read_excel('path_to_file.xls', 'Sheet1', index_col=None, na_values=['NA'])

added top-level function read_sql that is equivalent to the following

from pandas.io.sql import read_frame
read_frame(....)

DataFrame.to_html and DataFrame.to_latex now accept a path for their first argument (GH3702)
Do not allow astypes on datetime64[ns] except to object, and timedelta64[ns] to object/int
(GH3425)
The behavior of datetime64 dtypes has changed with respect to certain so-called reduction operations
(GH3726). The following operations now raise a TypeError when perfomed on a Series and return an
empty Series when performed on a DataFrame similar to performing these operations on, for example, a
DataFrame of slice objects:
sum, prod, mean, std, var, skew, kurt, corr, and cov
162

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

read_html now defaults to None when reading, and falls back on bs4 + html5lib when lxml fails to
parse. a list of parsers to try until success is also valid
The internal pandas class hierarchy has changed (slightly). The previous PandasObject now is called
PandasContainer and a new PandasObject has become the baseclass for PandasContainer as well
as Index, Categorical, GroupBy, SparseList, and SparseArray (+ their base classes). Currently,
PandasObject provides string methods (from StringMixin). (GH4090, GH4092)
New StringMixin that, given a __unicode__ method, gets python 2 and python 3 compatible string
methods (__str__, __bytes__, and __repr__). Plus string safety throughout. Now employed in many
places throughout the pandas library. (GH4090, GH4092)

1.12.2 I/O Enhancements

pd.read_html() can now parse HTML strings, files or urls and return DataFrames, courtesy of @cpcloud.
(GH3477, GH3605, GH3606, GH3616). It works with a single parser backend: BeautifulSoup4 + html5lib See
the docs
You can use pd.read_html() to read the output from DataFrame.to_html() like so
In [15]: df = DataFrame({'a': range(3), 'b': list('abc')})
In [16]: print(df)
a b
0 0 a
1 1 b
2 2 c
[3 rows x 2 columns]
In [17]: html = df.to_html()
In [18]: alist = pd.read_html(html, index_col=0)
In [19]:
a
0 True
1 True
2 True

print(df == alist[0])
b
True
True
True

[3 rows x 2 columns]

Note that alist here is a Python list so pd.read_html() and DataFrame.to_html() are not inverses.
pd.read_html() no longer performs hard conversion of date strings (GH3656).
Warning: You may have to install an older version of BeautifulSoup4, See the installation docs
Added module for reading and writing Stata files: pandas.io.stata (GH1512) accessable via
read_stata top-level function for reading, and to_stata DataFrame method for writing, See the docs
Added module for reading and writing json format files: pandas.io.json accessable via read_json toplevel function for reading, and to_json DataFrame method for writing, See the docs various issues (GH1226,
GH3804, GH3876, GH3867, GH1305)
MultiIndex column support for reading and writing csv format files

1.12. v0.12.0 (July 24, 2013)

163

pandas: powerful Python data analysis toolkit, Release 0.17.0

The header option in read_csv now accepts a list of the rows from which to read the index.
The option, tupleize_cols can now be specified in both to_csv and read_csv, to provide compatiblity for the pre 0.12 behavior of writing and reading MultIndex columns via a list of tuples. The
default in 0.12 is to write lists of tuples and not interpret list of tuples as a MultiIndex column.
Note: The default behavior in 0.12 remains unchanged from prior versions, but starting with 0.13, the
default to write and read MultiIndex columns will be in the new format. (GH3571, GH1651, GH3141)
If an index_col is not specified (e.g. you dont have an index, or wrote it with df.to_csv(...,
index=False), then any names on the columns index will be lost.
In [20]: from pandas.util.testing import makeCustomDataframe as mkdf
In [21]: df = mkdf(5,3,r_idx_nlevels=2,c_idx_nlevels=4)
In [22]: df.to_csv('mi.csv',tupleize_cols=False)
In [23]: print(open('mi.csv').read())
C0,,C_l0_g0,C_l0_g1,C_l0_g2
C1,,C_l1_g0,C_l1_g1,C_l1_g2
C2,,C_l2_g0,C_l2_g1,C_l2_g2
C3,,C_l3_g0,C_l3_g1,C_l3_g2
R0,R1,,,
R_l0_g0,R_l1_g0,R0C0,R0C1,R0C2
R_l0_g1,R_l1_g1,R1C0,R1C1,R1C2
R_l0_g2,R_l1_g2,R2C0,R2C1,R2C2
R_l0_g3,R_l1_g3,R3C0,R3C1,R3C2
R_l0_g4,R_l1_g4,R4C0,R4C1,R4C2

In [24]: pd.read_csv('mi.csv',header=[0,1,2,3],index_col=[0,1],tupleize_cols=False)
Out[24]:
C0
C_l0_g0 C_l0_g1 C_l0_g2
C1
C_l1_g0 C_l1_g1 C_l1_g2
C2
C_l2_g0 C_l2_g1 C_l2_g2
C3
C_l3_g0 C_l3_g1 C_l3_g2
R0
R1
R_l0_g0 R_l1_g0
R0C0
R0C1
R0C2
R_l0_g1 R_l1_g1
R1C0
R1C1
R1C2
R_l0_g2 R_l1_g2
R2C0
R2C1
R2C2
R_l0_g3 R_l1_g3
R3C0
R3C1
R3C2
R_l0_g4 R_l1_g4
R4C0
R4C1
R4C2
[5 rows x 3 columns]

Support for HDFStore (via PyTables 3.0.0) on Python3

Iterator support via read_hdf that automatically opens and closes the store when iteration is finished. This is
only for tables
In [25]: path = 'store_iterator.h5'
In [26]: DataFrame(randn(10,2)).to_hdf(path,'df',table=True)
In [27]: for df in read_hdf(path,'df', chunksize=3):
....:
print df
....:
0
1
0 0.713216 -0.778461

164

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

1 -0.661062
2 0.344342
0
3 -0.626968
4 -0.930687
5 0.949965
0
6 -0.402985
7 -0.241527
8 0.049355
0
9 -1.502767

0.862877
0.149565
1
-0.875772
-0.218983
-0.442354
1
1.111358
-0.670477
0.632633
1
-1.225492

read_csv will now throw a more informative error message when a file contains no columns, e.g., all newline
characters

1.12.3 Other Enhancements

DataFrame.replace() now allows regular expressions on contained Series with object dtype. See the
examples section in the regular docs Replacing via String Expression
For example you can do
In [25]: df = DataFrame({'a': list('ab..'), 'b': [1, 2, 3, 4]})
In [26]: df.replace(regex=r'\s*\.\s*', value=np.nan)
Out[26]:
a b
0
a 1
1
b 2
2 NaN 3
3 NaN 4
[4 rows x 2 columns]

to replace all occurrences of the string . with zero or more instances of surrounding whitespace with NaN.
Regular string replacement still works as expected. For example, you can do
In [27]: df.replace('.', np.nan)
Out[27]:
a b
0
a 1
1
b 2
2 NaN 3
3 NaN 4
[4 rows x 2 columns]

to replace all occurrences of the string . with NaN.

pd.melt() now accepts the optional parameters var_name and value_name to specify custom column
names of the returned DataFrame.
pd.set_option() now allows N option, value pairs (GH3667).
Lets say that we had an option a.b and another option b.c. We can set them at the same
time:

1.12. v0.12.0 (July 24, 2013)

165

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [28]: pd.get_option('a.b')
Out[28]: 2
In [29]: pd.get_option('b.c')
Out[29]: 3
In [30]: pd.set_option('a.b', 1, 'b.c', 4)
In [31]: pd.get_option('a.b')
Out[31]: 1
In [32]: pd.get_option('b.c')
Out[32]: 4

The filter method for group objects returns a subset of the original object. Suppose we want to take only
elements that belong to groups with a group sum greater than 2.
In [33]: sf = Series([1, 1, 2, 3, 3, 3])
In [34]: sf.groupby(sf).filter(lambda x: x.sum() > 2)
Out[34]:
3
3
4
3
5
3
dtype: int64

The argument of filter must a function that, applied to the group as a whole, returns True or False.
Another useful operation is filtering out elements that belong to groups with only a couple members.
In [35]: dff = DataFrame({'A': np.arange(8), 'B': list('aabbbbcc')})
In [36]: dff.groupby('B').filter(lambda x: len(x) > 2)
Out[36]:
A B
2 2 b
3 3 b
4 4 b
5 5 b
[4 rows x 2 columns]

Alternatively, instead of dropping the offending groups, we can return a like-indexed objects where the groups
that do not pass the filter are filled with NaNs.
In [37]: dff.groupby('B').filter(lambda x: len(x) > 2, dropna=False)
Out[37]:
A
B
0 NaN NaN
1 NaN NaN
2
2
b
3
3
b
4
4
b
5
5
b
6 NaN NaN
7 NaN NaN
[8 rows x 2 columns]

Series and DataFrame hist methods now take a figsize argument (GH3834)
166

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

DatetimeIndexes no longer try to convert mixed-integer indexes during join operations (GH3877)
Timestamp.min and Timestamp.max now represent valid Timestamp instances instead of the default datetime.min and datetime.max (respectively), thanks @SleepingPills
read_html now raises when no tables are found and BeautifulSoup==4.2.0 is detected (GH4214)

1.12.4 Experimental Features

Added experimental CustomBusinessDay class to support DateOffsets with custom holiday calendars
and custom weekmasks. (GH2301)
Note: This uses the numpy.busdaycalendar API introduced in Numpy 1.7 and therefore requires Numpy
1.7.0 or newer.
In [38]: from pandas.tseries.offsets import CustomBusinessDay
In [39]: from datetime import datetime
# As an interesting example, let's look at Egypt where
# a Friday-Saturday weekend is observed.
In [40]: weekmask_egypt = 'Sun Mon Tue Wed Thu'
# They also observe International Workers' Day so let's
# add that for a couple of years
In [41]: holidays = ['2012-05-01', datetime(2013, 5, 1), np.datetime64('2014-05-01')]
In [42]: bday_egypt = CustomBusinessDay(holidays=holidays, weekmask=weekmask_egypt)
In [43]: dt = datetime(2013, 4, 30)
In [44]: print(dt + 2 * bday_egypt)
2013-05-05 00:00:00
In [45]: dts = date_range(dt, periods=5, freq=bday_egypt)
In [46]: print(Series(dts.weekday, dts).map(Series('Mon Tue Wed Thu Fri Sat Sun'.split())))
2013-04-30
Tue
2013-05-02
Thu
2013-05-05
Sun
2013-05-06
Mon
2013-05-07
Tue
Freq: C, dtype: object

1.12.5 Bug Fixes

Plotting functions now raise a TypeError before trying to plot anything if the associated objects have have a
dtype of object (GH1818, GH3572, GH3911, GH3912), but they will try to convert object arrays to numeric
arrays if possible so that you can still plot, for example, an object array with floats. This happens before any
drawing takes place which elimnates any spurious plots from showing up.
fillna methods now raise a TypeError if the value parameter is a list or tuple.
Series.str now supports iteration (GH3638). You can iterate over the individual elements of each string in
the Series. Each iteration yields yields a Series with either a single character at each index of the original
Series or NaN. For example,

1.12. v0.12.0 (July 24, 2013)

167

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [47]: strs = 'go', 'bow', 'joe', 'slow'

In [48]: ds = Series(strs)
In [49]: for s in ds.str:
....:
print(s)
....:
0
g
1
b
2
j
3
s
dtype: object
0
o
1
o
2
o
3
l
dtype: object
0
NaN
1
w
2
e
3
o
dtype: object
0
NaN
1
NaN
2
NaN
3
w
dtype: object
In [50]: s
Out[50]:
0
NaN
1
NaN
2
NaN
3
w
dtype: object
In [51]: s.dropna().values.item() == 'w'
Out[51]: True

The last element yielded by the iterator will be a Series containing the last element of the longest string in
the Series with all other elements being NaN. Here since slow is the longest string and there are no other
strings with the same length w is the only non-null string in the yielded Series.
HDFStore
will retain index attributes (freq,tz,name) on recreation (GH3499)
will warn with a AttributeConflictWarning if you are attempting to append an index with a
different frequency than the existing, or attempting to append an index with a different name than the
existing
support datelike columns with a timezone as data_columns (GH2852)
Non-unique index support clarified (GH3468).
Fix assigning a new index to a duplicate index in a DataFrame would fail (GH3468)
Fix construction of a DataFrame with a duplicate index
ref_locs support to allow duplicative indices across dtypes, allows iget support to always find the index
(even across dtypes) (GH2194)
168

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

applymap on a DataFrame with a non-unique index now works (removed warning) (GH2786), and fix
(GH3230)
Fix to_csv to handle non-unique columns (GH3495)
Duplicate indexes with getitem will return items in the correct order (GH3455, GH3457) and handle missing elements like unique indices (GH3561)
Duplicate indexes with and empty DataFrame.from_records will return a correct frame (GH3562)
Concat to produce a non-unique columns when duplicates are across dtypes is fixed (GH3602)
Allow insert/delete to non-unique columns (GH3679)
Non-unique indexing with a slice via loc and friends fixed (GH3659)
Allow insert/delete to non-unique columns (GH3679)
Extend reindex to correctly deal with non-unique indices (GH3679)
DataFrame.itertuples() now works with frames with duplicate column names (GH3873)
Bug in non-unique indexing via iloc (GH4017); added takeable argument to reindex for locationbased taking
Allow non-unique indexing in series via .ix/.loc and __getitem__ (GH4246)
Fixed non-unique indexing memory allocation issue with .ix/.loc (GH4280)
DataFrame.from_records did not accept empty recarrays (GH3682)
read_html now correctly skips tests (GH3741)
Fixed a bug where DataFrame.replace with a compiled regular expression in the to_replace argument
wasnt working (GH3907)
Improved network test decorator to catch IOError (and therefore URLError as well). Added
with_connectivity_check decorator to allow explicitly checking a website as a proxy for seeing if there
is network connectivity. Plus, new optional_args decorator factory for decorators. (GH3910, GH3914)
Fixed testing issue where too many sockets where open thus leading to a connection reset issue (GH3982,
GH3985, GH4028, GH4054)
Fixed failing tests in test_yahoo, test_google where symbols were not retrieved but were being accessed
(GH3982, GH3985, GH4028, GH4054)
Series.hist will now take the figure from the current environment if one is not passed
Fixed bug where a 1xN DataFrame would barf on a 1xN mask (GH4071)
Fixed running of tox under python3 where the pickle import was getting rewritten in an incompatible way
(GH4062, GH4063)
Fixed bug where sharex and sharey were not being passed to grouped_hist (GH4089)
Fixed bug in DataFrame.replace where a nested dict wasnt being iterated over when regex=False
(GH4115)
Fixed bug in the parsing of microseconds when using the format argument in to_datetime (GH4152)
Fixed bug in PandasAutoDateLocator
MilliSecondLocator (GH3990)

where

invert_xaxis

triggered

incorrectly

Fixed bug in plotting that wasnt raising on invalid colormap for matplotlib 1.1.1 (GH4215)
Fixed the legend displaying in DataFrame.plot(kind=kde) (GH4216)
Fixed bug where Index slices werent carrying the name attribute (GH4226)

1.12. v0.12.0 (July 24, 2013)

169

pandas: powerful Python data analysis toolkit, Release 0.17.0

Fixed bug in initializing DatetimeIndex with an array of strings in a certain time zone (GH4229)
Fixed bug where html5lib wasnt being properly skipped (GH4265)
Fixed bug where get_data_famafrench wasnt using the correct file edges (GH4281)
See the full release notes or issue tracker on GitHub for a complete list.

1.13 v0.11.0 (April 22, 2013)

This is a major release from 0.10.1 and includes many new features and enhancements along with a large number of
bug fixes. The methods of Selecting Data have had quite a number of additions, and Dtype support is now full-fledged.
There are also a number of important API changes that long-time pandas users should pay close attention to.
There is a new section in the documentation, 10 Minutes to Pandas, primarily geared to new users.
There is a new section in the documentation, Cookbook, a collection of useful recipes in pandas (and that we want
contributions!).
There are several libraries that are now Recommended Dependencies

1.13.1 Selection Choices

Starting in 0.11.0, object selection has had a number of user-requested additions in order to support more explicit
location based indexing. Pandas now supports three types of multi-axis indexing.
.loc is strictly label based, will raise KeyError when the items are not found, allowed inputs are:
A single label, e.g. 5 or a, (note that 5 is interpreted as a label of the index. This use is not an integer
position along the index)
A list or array of labels [a, b, c]
A slice object with labels a:f, (note that contrary to usual python slices, both the start and the stop
are included!)
A boolean array
See more at Selection by Label
.iloc is strictly integer position based (from 0 to length-1 of the axis), will raise IndexError when the
requested indicies are out of bounds. Allowed inputs are:
An integer e.g. 5
A list or array of integers [4, 3, 0]
A slice object with ints 1:7
A boolean array
See more at Selection by Position
.ix supports mixed integer and label based access. It is primarily label based, but will fallback to integer
positional access. .ix is the most general and will support any of the inputs to .loc and .iloc, as well as
support for floating point label schemes. .ix is especially useful when dealing with mixed positional and label
based hierarchial indexes.
As using integer slices with .ix have different behavior depending on whether the slice is interpreted as position
based or label based, its usually better to be explicit and use .iloc or .loc.
See more at Advanced Indexing and Advanced Hierarchical.
170

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.13.2 Selection Deprecations

Starting in version 0.11.0, these methods may be deprecated in future versions.
irow
icol
iget_value
See the section Selection by Position for substitutes.

1.13.3 Dtypes
Numeric dtypes will propagate and can coexist in DataFrames. If a dtype is passed (either directly via the dtype
keyword, a passed ndarray, or a passed Series, then it will be preserved in DataFrame operations. Furthermore,
different numeric dtypes will NOT be combined. The following example will give you a taste.
In [1]: df1 = DataFrame(randn(8, 1), columns = ['A'], dtype = 'float32')
In [2]: df1
Out[2]:
A
0 1.392665
1 -0.123497
2 -0.402761
3 -0.246604
4 -0.288433
5 -0.763434
6 2.069526
7 -1.203569
[8 rows x 1 columns]
In [3]: df1.dtypes
Out[3]:
A
float32
dtype: object
In [4]: df2 = DataFrame(dict( A = Series(randn(8),dtype='float16'),
...:
B = Series(randn(8)),
...:
C = Series(randn(8),dtype='uint8') ))
...:
In [5]: df2
Out[5]:
A
B
0 0.591797 -0.038605
1 0.841309 -0.460478
2 -0.500977 -0.310458
3 -0.816406 0.866493
4 -0.207031 0.245972
5 -0.664062 0.319442
6 0.580566 1.378512
7 -0.965820 0.292502

C
0
1
0
254
0
1
1
255

[8 rows x 3 columns]
In [6]: df2.dtypes

1.13. v0.11.0 (April 22, 2013)

171

pandas: powerful Python data analysis toolkit, Release 0.17.0

Out[6]:
A
float16
B
float64
C
uint8
dtype: object
# here you get some upcasting
In [7]: df3 = df1.reindex_like(df2).fillna(value=0.0) + df2
In [8]: df3
Out[8]:
A
B
0 1.984462 -0.038605
1 0.717812 -0.460478
2 -0.903737 -0.310458
3 -1.063011 0.866493
4 -0.495465 0.245972
5 -1.427497 0.319442
6 2.650092 1.378512
7 -2.169390 0.292502

C
0
1
0
254
0
1
1
255

[8 rows x 3 columns]
In [9]: df3.dtypes
Out[9]:
A
float32
B
float64
C
float64
dtype: object

1.13.4 Dtype Conversion

This is lower-common-denomicator upcasting, meaning you get the dtype which can accomodate all of the types
In [10]: df3.values.dtype
Out[10]: dtype('float64')

Conversion
In [11]: df3.astype('float32').dtypes
Out[11]:
A
float32
B
float32
C
float32
dtype: object

Mixed Conversion
In [12]: df3['D'] = '1.'
In [13]: df3['E'] = '1'
In [14]: df3.convert_objects(convert_numeric=True).dtypes
Out[14]:
A
float32
B
float64
C
float64

172

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

D
float64
E
int64
dtype: object
# same, but specific dtype conversion
In [15]: df3['D'] = df3['D'].astype('float16')
In [16]: df3['E'] = df3['E'].astype('int32')
In [17]: df3.dtypes
Out[17]:
A
float32
B
float64
C
float64
D
float16
E
int32
dtype: object

Forcing Date coercion (and setting NaT when not datelike)

In [18]: from datetime import datetime
In [19]: s = Series([datetime(2001,1,1,0,0), 'foo', 1.0, 1,
....:
Timestamp('20010104'), '20010105'],dtype='O')
....:
In [20]: s.convert_objects(convert_dates='coerce')
Out[20]:
0
2001-01-01
1
NaT
2
NaT
3
NaT
4
2001-01-04
5
2001-01-05
dtype: datetime64[ns]

1.13.5 Dtype Gotchas

Platform Gotchas
Starting in 0.11.0, construction of DataFrame/Series will use default dtypes of int64 and float64, regardless
of platform. This is not an apparent change from earlier versions of pandas. If you specify dtypes, they WILL be
respected, however (GH2837)
The following will all result in int64 dtypes
In [21]: DataFrame([1,2],columns=['a']).dtypes
Out[21]:
a
int64
dtype: object
In [22]: DataFrame({'a' : [1,2] }).dtypes
Out[22]:
a
int64
dtype: object
In [23]: DataFrame({'a' : 1 }, index=range(2)).dtypes
Out[23]:

1.13. v0.11.0 (April 22, 2013)

173

pandas: powerful Python data analysis toolkit, Release 0.17.0

a
int64
dtype: object

Keep in mind that DataFrame(np.array([1,2])) WILL result in int32 on 32-bit platforms!

Upcasting Gotchas
Performing indexing operations on integer type data can easily upcast the data. The dtype of the input data will be
preserved in cases where nans are not introduced.
In [24]: dfi = df3.astype('int32')
In [25]: dfi['D'] = dfi['D'].astype('int64')
In [26]: dfi
Out[26]:
A B
C
0 1 0
0
1 0 0
1
2 0 0
0
3 -1 0 254
4 0 0
0
5 -1 0
1
6 2 1
1
7 -2 0 255

D
1
1
1
1
1
1
1
1

E
1
1
1
1
1
1
1
1

[8 rows x 5 columns]
In [27]: dfi.dtypes
Out[27]:
A
int32
B
int32
C
int32
D
int64
E
int32
dtype: object
In [28]: casted = dfi[dfi>0]
In [29]: casted
Out[29]:
A
B
C
0
1 NaN NaN
1 NaN NaN
1
2 NaN NaN NaN
3 NaN NaN 254
4 NaN NaN NaN
5 NaN NaN
1
6
2
1
1
7 NaN NaN 255

D
1
1
1
1
1
1
1
1

E
1
1
1
1
1
1
1
1

[8 rows x 5 columns]
In [30]: casted.dtypes
Out[30]:
A
float64
B
float64
C
float64
D
int64

174

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

E
int32
dtype: object

While float dtypes are unchanged.

In [31]: df4 = df3.copy()
In [32]: df4['A'] = df4['A'].astype('float32')
In [33]: df4.dtypes
Out[33]:
A
float32
B
float64
C
float64
D
float16
E
int32
dtype: object
In [34]: casted = df4[df4>0]
In [35]: casted
Out[35]:
A
B
0 1.984462
NaN
1 0.717812
NaN
2
NaN
NaN
3
NaN 0.866493
4
NaN 0.245972
5
NaN 0.319442
6 2.650092 1.378512
7
NaN 0.292502

C
NaN
1
NaN
254
NaN
1
1
255

D
1
1
1
1
1
1
1
1

E
1
1
1
1
1
1
1
1

[8 rows x 5 columns]
In [36]: casted.dtypes
Out[36]:
A
float32
B
float64
C
float64
D
float16
E
int32
dtype: object

1.13.6 Datetimes Conversion

Datetime64[ns] columns in a DataFrame (or a Series) allow the use of np.nan to indicate a nan value, in addition to the traditional NaT, or not-a-time. This allows convenient nan setting in a generic way. Furthermore
datetime64[ns] columns are created by default, when passed datetimelike objects (this change was introduced in
0.10.1) (GH2809, GH2810)
In [37]: df = DataFrame(randn(6,2),date_range('20010102',periods=6),columns=['A','B'])
In [38]: df['timestamp'] = Timestamp('20010103')
In [39]: df
Out[39]:
A

1.13. v0.11.0 (April 22, 2013)

timestamp

175

pandas: powerful Python data analysis toolkit, Release 0.17.0

2001-01-02
2001-01-03
2001-01-04
2001-01-05
2001-01-06
2001-01-07

1.023958
1.236475
-0.270630
-0.440747
-0.632102
-1.444787

0.660103
-2.170629
-1.685677
-0.115070
-0.585977
-0.201135

2001-01-03
2001-01-03
2001-01-03
2001-01-03
2001-01-03
2001-01-03

[6 rows x 3 columns]
# datetime64[ns] out of the box
In [40]: df.get_dtype_counts()
Out[40]:
datetime64[ns]
1
float64
2
dtype: int64
# use the traditional nan, which is mapped to NaT internally
In [41]: df.ix[2:4,['A','timestamp']] = np.nan
In [42]: df
Out[42]:
A
B
2001-01-02 1.023958 0.660103
2001-01-03 1.236475 -2.170629
2001-01-04
NaN -1.685677
2001-01-05
NaN -0.115070
2001-01-06 -0.632102 -0.585977
2001-01-07 -1.444787 -0.201135

timestamp
2001-01-03
2001-01-03
NaT
NaT
2001-01-03
2001-01-03

[6 rows x 3 columns]

Astype conversion on datetime64[ns] to object, implicity converts NaT to np.nan

In [43]: import datetime
In [44]: s = Series([datetime.datetime(2001, 1, 2, 0, 0) for i in range(3)])
In [45]: s.dtype
Out[45]: dtype('<M8[ns]')
In [46]: s[1] = np.nan
In [47]: s
Out[47]:
0
2001-01-02
1
NaT
2
2001-01-02
dtype: datetime64[ns]
In [48]: s.dtype
Out[48]: dtype('<M8[ns]')
In [49]: s = s.astype('O')
In [50]: s
Out[50]:
0
2001-01-02 00:00:00
1
NaT

176

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

2
2001-01-02 00:00:00
dtype: object
In [51]: s.dtype
Out[51]: dtype('O')

1.13.7 API changes

Added to_series() method to indicies, to facilitate the creation of indexers (GH3275)
HDFStore
added the method select_column to select a single column from a table as a Series.
deprecated the unique method, can be replicated by select_column(key,column).unique()
min_itemsize parameter to append will now automatically create data_columns for passed keys

1.13.8 Enhancements
Improved performance of df.to_csv() by up to 10x in some cases. (GH3059)
Numexpr is now a Recommended Dependencies, to accelerate certain types of numerical and boolean operations
Bottleneck is now a Recommended Dependencies, to accelerate certain types of nan operations
HDFStore
support read_hdf/to_hdf API similar to read_csv/to_csv
In [52]: df = DataFrame(dict(A=lrange(5), B=lrange(5)))
In [53]: df.to_hdf('store.h5','table',append=True)
In [54]: read_hdf('store.h5', 'table', where = ['index>2'])
Out[54]:
A B
3 3 3
4 4 4
[2 rows x 2 columns]

provide dotted attribute access to get from stores, e.g. store.df == store[df]
new keywords iterator=boolean, and chunksize=number_in_a_chunk are provided to support iteration on select and select_as_multiple (GH3076)
You can now select timestamps from an unordered timeseries similarly to an ordered timeseries (GH2437)
You can now select with a string from a DataFrame with a datelike index, in a similar way to a Series (GH3070)
In [55]: idx = date_range("2001-10-1", periods=5, freq='M')
In [56]: ts = Series(np.random.rand(len(idx)),index=idx)
In [57]: ts['2001']
Out[57]:
2001-10-31
0.663256
2001-11-30
0.079126
2001-12-31
0.587699

1.13. v0.11.0 (April 22, 2013)

177

pandas: powerful Python data analysis toolkit, Release 0.17.0

Freq: M, dtype: float64

In [58]: df = DataFrame(dict(A = ts))
In [59]: df['2001']
Out[59]:
A
2001-10-31 0.663256
2001-11-30 0.079126
2001-12-31 0.587699
[3 rows x 1 columns]

Squeeze to possibly remove length 1 dimensions from an object.

In [60]: p = Panel(randn(3,4,4),items=['ItemA','ItemB','ItemC'],
....:
major_axis=date_range('20010102',periods=4),
....:
minor_axis=['A','B','C','D'])
....:
In [61]: p
Out[61]:
<class 'pandas.core.panel.Panel'>
Dimensions: 3 (items) x 4 (major_axis) x 4 (minor_axis)
Items axis: ItemA to ItemC
Major_axis axis: 2001-01-02 00:00:00 to 2001-01-05 00:00:00
Minor_axis axis: A to D
In [62]: p.reindex(items=['ItemA']).squeeze()
Out[62]:
A
B
C
D
2001-01-02 -1.203403 0.425882 -0.436045 -0.982462
2001-01-03 0.348090 -0.969649 0.121731 0.202798
2001-01-04 1.215695 -0.218549 -0.631381 -0.337116
2001-01-05 0.404238 0.907213 -0.865657 0.483186
[4 rows x 4 columns]
In [63]: p.reindex(items=['ItemA'],minor=['B']).squeeze()
Out[63]:
2001-01-02
0.425882
2001-01-03
-0.969649
2001-01-04
-0.218549
2001-01-05
0.907213
Freq: D, Name: B, dtype: float64

In pd.io.data.Options,
Fix bug when trying to fetch data for the current month when already past expiry.
Now using lxml to scrape html instead of BeautifulSoup (lxml was faster).
New instance variables for calls and puts are automatically created when a method that creates them is
called. This works for current month where the instance variables are simply calls and puts. Also
works for future expiry months and save the instance variable as callsMMYY or putsMMYY, where
MMYY are, respectively, the month and year of the options expiry.
Options.get_near_stock_price now allows the user to specify the month for which to get relevant options data.

178

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Options.get_forward_data now has optional kwargs near and above_below. This allows the
user to specify if they would like to only return forward looking data for options near the current stock
price. This just obtains the data from Options.get_near_stock_price instead of Options.get_xxx_data()
(GH2758).
Cursor coordinate information is now displayed in time-series plots.
added option display.max_seq_items to control the number of elements printed per sequence pprinting it.
(GH2979)
added option display.chop_threshold to control display of small numerical values. (GH2739)
added option display.max_info_rows to prevent verbose_info from being calculated for frames above 1M rows
(configurable). (GH2807, GH2918)
value_counts() now accepts a normalize argument, for normalized histograms. (GH2710).
DataFrame.from_records now accepts not only dicts but any instance of the collections.Mapping ABC.
added option display.mpl_style providing a
https://gist.github.com/huyng/816622 (GH3075).

sleeker

visual

style

for

plots.

Based

Treat boolean values as integers (values 1 and 0) for numeric operations. (GH2641)
to_html() now accepts an optional escape argument to control reserved HTML character escaping (enabled
by default) and escapes &, in addition to < and >. (GH2919)
See the full release notes or issue tracker on GitHub for a complete list.

1.14 v0.10.1 (January 22, 2013)

This is a minor release from 0.10.0 and includes new features, enhancements, and bug fixes. In particular, there is
substantial new HDFStore functionality contributed by Jeff Reback.
An undesired API breakage with functions taking the inplace option has been reverted and deprecation warnings
added.

1.14.1 API changes

Functions taking an inplace option return the calling object as before. A deprecation message has been added
Groupby aggregations Max/Min no longer exclude non-numeric data (GH2700)
Resampling an empty DataFrame now returns an empty DataFrame instead of raising an exception (GH2640)
The file reader will now raise an exception when NA values are found in an explicitly specified integer column
instead of converting the column to float (GH2631)
DatetimeIndex.unique now returns a DatetimeIndex with the same name and
timezone instead of an array (GH2563)

1.14.2 New features

MySQL support for database (contribution from Dan Allan)

1.14. v0.10.1 (January 22, 2013)

179

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.14.3 HDFStore
You may need to upgrade your existing data files. Please visit the compatibility section in the main docs.
You can designate (and index) certain columns that you want to be able to perform queries on a table, by passing a list
to data_columns
In [1]: store = HDFStore('store.h5')
In [2]: df = DataFrame(randn(8, 3), index=date_range('1/1/2000', periods=8),
...:
columns=['A', 'B', 'C'])
...:
In [3]: df['string'] = 'foo'
In [4]: df.ix[4:6,'string'] = np.nan
In [5]: df.ix[7:9,'string'] = 'bar'
In [6]: df['string2'] = 'cool'
In [7]: df
Out[7]:
2000-01-01
2000-01-02
2000-01-03
2000-01-04
2000-01-05
2000-01-06
2000-01-07
2000-01-08

A
1.885136
0.180759
-0.294467
3.127110
-0.242846
0.820521
-0.034086
-2.290958

B
C string string2
-0.183873 2.550850
foo
cool
-1.117089 0.061462
foo
cool
-0.591411 -0.876691
foo
cool
1.451130 0.045152
foo
cool
1.195819 1.533294
NaN
cool
-0.281201 1.651561
NaN
cool
0.252394 -0.498772
foo
cool
-1.601262 -0.256718
bar
cool

[8 rows x 5 columns]
# on-disk operations
In [8]: store.append('df', df, data_columns = ['B','C','string','string2'])
In [9]: store.select('df',[ 'B > 0', 'string == foo' ])
Out[9]:
Empty DataFrame
Columns: [A, B, C, string, string2]
Index: []
[0 rows x 5 columns]
# this is in-memory version of this type of selection
In [10]: df[(df.B > 0) & (df.string == 'foo')]
Out[10]:
A
B
C string string2
2000-01-04 3.127110 1.451130 0.045152
foo
cool
2000-01-07 -0.034086 0.252394 -0.498772
foo
cool
[2 rows x 5 columns]

Retrieving unique values in an indexable or data column.

# note that this is deprecated as of 0.14.0
# can be replicated by: store.select_column('df','index').unique()

180

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

store.unique('df','index')
store.unique('df','string')

You can now store datetime64 in data columns

In [11]: df_mixed

= df.copy()

In [12]: df_mixed['datetime64'] = Timestamp('20010102')

In [13]: df_mixed.ix[3:4,['A','B']] = np.nan
In [14]: store.append('df_mixed', df_mixed)
In [15]: df_mixed1 = store.select('df_mixed')
In [16]: df_mixed1
Out[16]:
2000-01-01
2000-01-02
2000-01-03
2000-01-04
2000-01-05
2000-01-06
2000-01-07
2000-01-08

A
1.885136
0.180759
-0.294467
NaN
-0.242846
0.820521
-0.034086
-2.290958

B
C string string2 datetime64
-0.183873 2.550850
foo
cool 2001-01-02
-1.117089 0.061462
foo
cool 2001-01-02
-0.591411 -0.876691
foo
cool 2001-01-02
NaN 0.045152
foo
cool 2001-01-02
1.195819 1.533294
NaN
cool 2001-01-02
-0.281201 1.651561
NaN
cool 2001-01-02
0.252394 -0.498772
foo
cool 2001-01-02
-1.601262 -0.256718
bar
cool 2001-01-02

[8 rows x 6 columns]
In [17]: df_mixed1.get_dtype_counts()
Out[17]:
datetime64[ns]
1
float64
3
object
2
dtype: int64

You can pass columns keyword to select to filter a list of the return columns, this is equivalent to passing a
Term(columns,list_of_columns_to_filter)
In [18]: store.select('df',columns = ['A','B'])
Out[18]:
A
B
2000-01-01 1.885136 -0.183873
2000-01-02 0.180759 -1.117089
2000-01-03 -0.294467 -0.591411
2000-01-04 3.127110 1.451130
2000-01-05 -0.242846 1.195819
2000-01-06 0.820521 -0.281201
2000-01-07 -0.034086 0.252394
2000-01-08 -2.290958 -1.601262
[8 rows x 2 columns]

HDFStore now serializes multi-index dataframes when appending tables.

In [19]: index = MultiIndex(levels=[['foo', 'bar', 'baz', 'qux'],
....:
['one', 'two', 'three']],
....:
labels=[[0, 0, 0, 1, 1, 2, 2, 3, 3, 3],
....:
[0, 1, 2, 0, 1, 1, 2, 0, 1, 2]],

1.14. v0.10.1 (January 22, 2013)

181

pandas: powerful Python data analysis toolkit, Release 0.17.0

names=['foo', 'bar'])

....:
....:

In [20]: df = DataFrame(np.random.randn(10, 3), index=index,

....:
columns=['A', 'B', 'C'])
....:
In [21]: df
Out[21]:
foo bar
foo one
two
three
bar one
two
baz two
three
qux one
two
three

0.239369
-1.948006
-0.361633
0.152288
-0.859278
-0.647604
-0.342928
0.104020
-0.019240
-0.637855

0.174122
0.980347
-0.761218
-0.862613
1.498195
1.511487
-0.007364
2.052171
-1.713238
0.215109

-1.131794
-0.674429
1.768215
-0.210968
0.462413
-0.727189
1.427674
-1.230963
0.838912
-1.515362

[10 rows x 3 columns]

In [22]: store.append('mi',df)
In [23]: store.select('mi')
Out[23]:
A
B
foo bar
foo one
0.239369 0.174122
two
-1.948006 0.980347
three -0.361633 -0.761218
bar one
0.152288 -0.862613
two
-0.859278 1.498195
baz two
-0.647604 1.511487
three -0.342928 -0.007364
qux one
0.104020 2.052171
two
-0.019240 -1.713238
three -0.637855 0.215109

C
-1.131794
-0.674429
1.768215
-0.210968
0.462413
-0.727189
1.427674
-1.230963
0.838912
-1.515362

[10 rows x 3 columns]

# the levels are automatically included as data columns
In [24]: store.select('mi', Term('foo=bar'))
Out[24]:
Empty DataFrame
Columns: [A, B, C]
Index: []
[0 rows x 3 columns]

Multi-table creation via append_to_multiple and selection via select_as_multiple can create/select from
multiple tables and return a combined result, by using where on a selector table.
In [25]: df_mt = DataFrame(randn(8, 6), index=date_range('1/1/2000', periods=8),
....:
columns=['A', 'B', 'C', 'D', 'E', 'F'])
....:

182

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [26]: df_mt['foo'] = 'bar'

# you can also create the tables individually

In [27]: store.append_to_multiple({ 'df1_mt' : ['A','B'], 'df2_mt' : None }, df_mt, selector = 'df1_m

In [28]: store
Out[28]:
<class 'pandas.io.pytables.HDFStore'>
File path: store.h5
/df
frame_table (typ->appendable,nrows->8,ncols->5,indexers->[index],dc->[B,C,strin
/df1_mt
frame_table (typ->appendable,nrows->8,ncols->2,indexers->[index],dc->[A,B])
/df2_mt
frame_table (typ->appendable,nrows->8,ncols->5,indexers->[index])
/df_mixed
frame_table (typ->appendable,nrows->8,ncols->6,indexers->[index])
/mi
frame_table (typ->appendable_multi,nrows->10,ncols->5,indexers->[index],dc->[ba
# indiviual tables were created
In [29]: store.select('df1_mt')
Out[29]:
A
B
2000-01-01 1.586924 -0.447974
2000-01-02 -0.102206 0.870302
2000-01-03 1.249874 1.458210
2000-01-04 -0.616293 0.150468
2000-01-05 -0.431163 0.016640
2000-01-06 0.800353 -0.451572
2000-01-07 1.239198 0.185437
2000-01-08 -0.040863 0.290110
[8 rows x 2 columns]
In [30]: store.select('df2_mt')
Out[30]:
C
D
E
F
2000-01-01 -1.573998 0.630925 -0.071659 -1.277640
2000-01-02 1.275280 -1.199212 1.060780 1.673018
2000-01-03 -0.710542 0.825392 1.557329 1.993441
2000-01-04 0.132104 0.580923 -0.128750 1.445964
2000-01-05 0.904578 -1.645852 -0.688741 0.228006
2000-01-06 0.831767 0.228760 0.932498 -2.200069
2000-01-07 -0.540770 -0.370038 1.298390 1.662964
2000-01-08 -0.096145 1.717830 -0.462446 -0.112019

foo
bar
bar
bar
bar
bar
bar
bar
bar

[8 rows x 5 columns]
# as a multiple
In [31]: store.select_as_multiple(['df1_mt','df2_mt'], where = [ 'A>0','B>0' ], selector = 'df1_mt')
Out[31]:
A
B
C
D
E
F foo
2000-01-03 1.249874 1.458210 -0.710542 0.825392 1.557329 1.993441 bar
2000-01-07 1.239198 0.185437 -0.540770 -0.370038 1.298390 1.662964 bar
[2 rows x 7 columns]

Enhancements
HDFStore now can read native PyTables table format tables
You can pass nan_rep = my_nan_rep to append, to change the default nan representation on disk
(which converts to/from np.nan), this defaults to nan.
1.14. v0.10.1 (January 22, 2013)

183

pandas: powerful Python data analysis toolkit, Release 0.17.0

You can pass index to append. This defaults to True. This will automagically create indicies on the
indexables and data columns of the table
You can pass chunksize=an integer to append, to change the writing chunksize (default is 50000).
This will signficantly lower your memory usage on writing.
You can pass expectedrows=an integer to the first append, to set the TOTAL number of expectedrows
that PyTables will expected. This will optimize read/write performance.
Select now supports passing start and stop to provide selection space limiting in selection.
Greatly improved ISO8601 (e.g., yyyy-mm-dd) date parsing for file parsers (GH2698)
Allow DataFrame.merge to handle combinatorial sizes too large for 64-bit integer (GH2690)
Series now has unary negation (-series) and inversion (~series) operators (GH2686)
DataFrame.plot now includes a logx parameter to change the x-axis to log scale (GH2327)
Series arithmetic operators can now handle constant and ndarray input (GH2574)
ExcelFile now takes a kind argument to specify the file type (GH2613)
A faster implementation for Series.str methods (GH2602)
Bug Fixes
HDFStore tables can now store float32 types correctly (cannot be mixed with float64 however)
Fixed Google Analytics prefix when specifying request segment (GH2713).
Function to reset Google Analytics token store so users can recover from improperly setup client secrets
(GH2687).
Fixed groupby bug resulting in segfault when passing in MultiIndex (GH2706)
Fixed bug where passing a Series with datetime64 values into to_datetime results in bogus output values
(GH2699)
Fixed bug in pattern in HDFStore expressions when pattern is not a valid regex (GH2694)
Fixed performance issues while aggregating boolean data (GH2692)
When given a boolean mask key and a Series of new values, Series __setitem__ will now align the incoming
values with the original Series (GH2686)
Fixed MemoryError caused by performing counting sort on sorting MultiIndex levels with a very large number
of combinatorial values (GH2684)
Fixed bug that causes plotting to fail when the index is a DatetimeIndex with a fixed-offset timezone (GH2683)
Corrected businessday subtraction logic when the offset is more than 5 bdays and the starting date is on a
weekend (GH2680)
Fixed C file parser behavior when the file has more columns than data (GH2668)
Fixed file reader bug that misaligned columns with data in the presence of an implicit column and a specified
usecols value
DataFrames with numerical or datetime indices are now sorted prior to plotting (GH2609)
Fixed DataFrame.from_records error when passed columns, index, but empty records (GH2633)
Several bug fixed for Series operations when dtype is datetime64 (GH2689, GH2629, GH2626)
See the full release notes or issue tracker on GitHub for a complete list.

184

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.15 v0.10.0 (December 17, 2012)

This is a major release from 0.9.1 and includes many new features and enhancements along with a large number of
bug fixes. There are also a number of important API changes that long-time pandas users should pay close attention
to.

1.15.1 File parsing new features

The delimited file parsing engine (the guts of read_csv and read_table) has been rewritten from the ground up
and now uses a fraction the amount of memory while parsing, while being 40% or more faster in most use cases (in
some cases much faster).
There are also many new features:
Much-improved Unicode handling via the encoding option.
Column filtering (usecols)
Dtype specification (dtype argument)
Ability to specify strings to be recognized as True/False
Ability to yield NumPy record arrays (as_recarray)
High performance delim_whitespace option
Decimal format (e.g. European format) specification
Easier CSV dialect options: escapechar, lineterminator, quotechar, etc.
More robust handling of many exceptional kinds of files observed in the wild

1.15.2 API changes

Deprecated DataFrame BINOP TimeSeries special case behavior
The default behavior of binary operations between a DataFrame and a Series has always been to align on the
DataFrames columns and broadcast down the rows, except in the special case that the DataFrame contains time
series. Since there are now method for each binary operator enabling you to specify how you want to broadcast, we
are phasing out this special case (Zen of Python: Special cases arent special enough to break the rules). Heres what
Im talking about:
In [1]: import pandas as pd
In [2]: df = pd.DataFrame(np.random.randn(6, 4),
...:
index=pd.date_range('1/1/2000', periods=6))
...:
In [3]: df
Out[3]:
2000-01-01
2000-01-02
2000-01-03
2000-01-04
2000-01-05
2000-01-06

0
-0.134024
-1.626124
-1.565401
-0.302741
0.268766
-0.639757

1
2
3
-0.205969 1.348944 -1.198246
0.982041 0.059493 -0.460111
-0.025706 0.942864 2.502156
0.261551 -0.066342 0.897097
-1.225092 0.582752 -1.490764
-0.952750 -0.892402 0.505987

[6 rows x 4 columns]

1.15. v0.10.0 (December 17, 2012)

185

pandas: powerful Python data analysis toolkit, Release 0.17.0

# deprecated now
In [4]: df - df[0]
Out[4]:
2000-01-01 00:00:00
2000-01-01
NaN
2000-01-02
NaN
2000-01-03
NaN
2000-01-04
NaN
2000-01-05
NaN
2000-01-06
NaN

2000-01-02 00:00:00
NaN
NaN
NaN
NaN
NaN
NaN

2000-01-03 00:00:00
NaN
NaN
NaN
NaN
NaN
NaN

2000-01-04 00:00:00
NaN
NaN
NaN
NaN
NaN
NaN

2000-01-05 00:00:00
NaN
NaN
NaN
NaN
NaN
NaN

2000-01-06 00:00:00
NaN
NaN
NaN
NaN
NaN
NaN

0
NaN
NaN
NaN
NaN
NaN
NaN

2000-01-01
2000-01-02
2000-01-03
2000-01-04
2000-01-05
2000-01-06

1
NaN
NaN
NaN
NaN
NaN
NaN

2
NaN
NaN
NaN
NaN
NaN
NaN

3
NaN
NaN
NaN
NaN
NaN
NaN

[6 rows x 10 columns]
# Change your code to
In [5]: df.sub(df[0], axis=0) # align on axis 0 (rows)
Out[5]:
0
1
2
3
2000-01-01 0 -0.071946 1.482967 -1.064223
2000-01-02 0 2.608165 1.685618 1.166013
2000-01-03 0 1.539695 2.508265 4.067556
2000-01-04 0 0.564293 0.236399 1.199839
2000-01-05 0 -1.493857 0.313986 -1.759530
2000-01-06 0 -0.312993 -0.252645 1.145744
[6 rows x 4 columns]

You will get a deprecation warning in the 0.10.x series, and the deprecated functionality will be removed in 0.11 or
later.
Altered resample default behavior
The default time series resample binning behavior of daily D and higher frequencies has been changed to
closed=left, label=left. Lower nfrequencies are unaffected. The prior defaults were causing a great
deal of confusion for users, especially resampling data to daily frequency (which labeled the aggregated group with
the end of the interval: the next day).
Note:
In [6]: dates = pd.date_range('1/1/2000', '1/5/2000', freq='4h')
In [7]: series = Series(np.arange(len(dates)), index=dates)
In [8]: series

186

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Out[8]:
2000-01-01
2000-01-01
2000-01-01
2000-01-01
2000-01-01
2000-01-01
2000-01-02

00:00:00
04:00:00
08:00:00
12:00:00
16:00:00
20:00:00
00:00:00

2000-01-04 00:00:00
2000-01-04 04:00:00
2000-01-04 08:00:00
2000-01-04 12:00:00
2000-01-04 16:00:00
2000-01-04 20:00:00
2000-01-05 00:00:00
Freq: 4H, dtype: int32

0
1
2
3
4
5
6
..
18
19
20
21
22
23
24

In [9]: series.resample('D', how='sum')

Out[9]:
2000-01-01
15
2000-01-02
51
2000-01-03
87
2000-01-04
123
2000-01-05
24
Freq: D, dtype: int32
# old behavior
In [10]: series.resample('D', how='sum', closed='right', label='right')
Out[10]:
2000-01-01
0
2000-01-02
21
2000-01-03
57
2000-01-04
93
2000-01-05
129
Freq: D, dtype: int32

Infinity and negative infinity are no longer treated as NA by isnull and notnull. That they every were was
a relic of early pandas. This behavior can be re-enabled globally by the mode.use_inf_as_null option:
In [11]: s = pd.Series([1.5, np.inf, 3.4, -np.inf])
In [12]: pd.isnull(s)
Out[12]:
0
False
1
False
2
False
3
False
dtype: bool
In [13]: s.fillna(0)
Out[13]:
0
1.500000
1
inf
2
3.400000
3
-inf
dtype: float64

1.15. v0.10.0 (December 17, 2012)

187

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [14]: pd.set_option('use_inf_as_null', True)

In [15]: pd.isnull(s)
Out[15]:
0
False
1
True
2
False
3
True
dtype: bool
In [16]: s.fillna(0)
Out[16]:
0
1.5
1
0.0
2
3.4
3
0.0
dtype: float64
In [17]: pd.reset_option('use_inf_as_null')

Methods with the inplace option now all return None instead of the calling object. E.g. code written like
df = df.fillna(0, inplace=True) may stop working. To fix, simply delete the unnecessary variable
assignment.
pandas.merge no longer sorts the group keys (sort=False) by default. This was done for performance
reasons: the group-key sorting is often one of the more expensive parts of the computation and is often unnecessary.
The default column names for a file with no header have been changed to the integers 0 through N - 1. This
is to create consistency with the DataFrame constructor with no columns specified. The v0.9.0 behavior (names
X0, X1, ...) can be reproduced by specifying prefix=X:
In [18]: data= 'a,b,c\n1,Yes,2\n3,No,4'
In [19]: print(data)
a,b,c
1,Yes,2
3,No,4
In [20]: pd.read_csv(StringIO(data), header=None)
Out[20]:
0
1 2
0 a
b c
1 1 Yes 2
2 3
No 4
[3 rows x 3 columns]
In [21]: pd.read_csv(StringIO(data), header=None, prefix='X')
Out[21]:
X0
X1 X2
0 a
b c
1 1 Yes 2
2 3
No 4
[3 rows x 3 columns]

Values like Yes and No are not interpreted as boolean by default, though this can be controlled by new
true_values and false_values arguments:
188

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [22]: print(data)
a,b,c
1,Yes,2
3,No,4
In [23]: pd.read_csv(StringIO(data))
Out[23]:
a
b c
0 1 Yes 2
1 3
No 4
[2 rows x 3 columns]
In [24]: pd.read_csv(StringIO(data), true_values=['Yes'], false_values=['No'])
Out[24]:
a
b c
0 1
True 2
1 3 False 4
[2 rows x 3 columns]

The file parsers will not recognize non-string values arising from a converter function as NA if passed in the
na_values argument. Its better to do post-processing using the replace function instead.
Calling fillna on Series or DataFrame with no arguments is no longer valid code. You must either specify a
fill value or an interpolation method:
In [25]: s = Series([np.nan, 1., 2., np.nan, 4])
In [26]: s
Out[26]:
0
NaN
1
1
2
2
3
NaN
4
4
dtype: float64
In [27]: s.fillna(0)
Out[27]:
0
0
1
1
2
2
3
0
4
4
dtype: float64
In [28]: s.fillna(method='pad')
Out[28]:
0
NaN
1
1
2
2
3
2
4
4
dtype: float64

Convenience methods ffill and bfill have been added:

1.15. v0.10.0 (December 17, 2012)

189

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [29]: s.ffill()
Out[29]:
0
NaN
1
1
2
2
3
2
4
4
dtype: float64

Series.apply will now operate on a returned value from the applied function, that is itself a series, and
possibly upcast the result to a DataFrame
In [30]: def f(x):
....:
return Series([ x, x**2 ], index = ['x', 'x^2'])
....:
In [31]: s = Series(np.random.rand(5))
In [32]: s
Out[32]:
0
0.717478
1
0.815199
2
0.452478
3
0.848385
4
0.235477
dtype: float64
In [33]: s.apply(f)
Out[33]:
x
x^2
0 0.717478 0.514775
1 0.815199 0.664550
2 0.452478 0.204737
3 0.848385 0.719757
4 0.235477 0.055449
[5 rows x 2 columns]

New API functions for working with pandas options (GH2097):

get_option / set_option - get/set the value of an option. Partial names are accepted. reset_option - reset one or more options to their default value. Partial names are accepted. describe_option - print a description of one or more options. When called with no arguments.
print all registered options.
Note: set_printoptions/ reset_printoptions are now deprecated (but functioning), the print options now live under display.XYZ. For example:
In [34]: get_option("display.max_rows")
Out[34]: 15

to_string() methods now always return unicode strings (GH2224).

190

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.15.3 New features

1.15.4 Wide DataFrame Printing
Instead of printing the summary information, pandas now splits the string representation across multiple rows by
default:
In [35]: wide_frame = DataFrame(randn(5, 16))
In [36]: wide_frame
Out[36]:
0
1
2
3
4
5
6
0 -0.681624 0.191356 1.180274 -0.834179 0.703043 0.166568 -0.583599
1 0.441522 -0.316864 -0.017062 1.570114 -0.360875 -0.880096 0.235532
2 -0.412451 -0.462580 0.422194 0.288403 -0.487393 -0.777639 0.055865
3 -0.277255 1.331263 0.585174 -0.568825 -0.719412 1.191340 -0.456362
4 -1.642511 0.432560 1.218080 -0.564705 -0.581790 0.286071 0.048725
7
8
9
10
11
12
13
0 -1.201796 -1.422811 -0.882554 1.209871 -0.941235 0.863067 -0.336232
1 0.207232 -1.983857 -1.702547 -1.621234 -0.906840 1.014601 -0.475108
2 1.383381 0.085638 0.246392 0.965887 0.246354 -0.727728 -0.094414
3 0.089931 0.776079 0.752889 -1.195795 -1.425911 -0.548829 0.774225
4 1.002440 1.276582 0.054399 0.241963 -0.471786 0.314510 -0.059986

0
1
2
3
4

14
15
-0.976847 0.033862
-0.358944 1.262942
-0.276854 0.158399
0.740501 1.510263
-2.069319 -1.115104

[5 rows x 16 columns]

The old behavior of printing out summary information can be achieved via the expand_frame_repr print option:
In [37]: pd.set_option('expand_frame_repr', False)

In [38]: wide_frame
Out[38]:
0
1
2
3
4
5
6
7
8
9
0 -0.681624 0.191356 1.180274 -0.834179 0.703043 0.166568 -0.583599 -1.201796 -1.422811 -0.882554
1 0.441522 -0.316864 -0.017062 1.570114 -0.360875 -0.880096 0.235532 0.207232 -1.983857 -1.702547
2 -0.412451 -0.462580 0.422194 0.288403 -0.487393 -0.777639 0.055865 1.383381 0.085638 0.246392
3 -0.277255 1.331263 0.585174 -0.568825 -0.719412 1.191340 -0.456362 0.089931 0.776079 0.752889
4 -1.642511 0.432560 1.218080 -0.564705 -0.581790 0.286071 0.048725 1.002440 1.276582 0.054399
[5 rows x 16 columns]

The width of each line can be changed via line_width (80 by default):
In [39]: pd.set_option('line_width', 40)
line_width has been deprecated, use display.width instead (currently both are
identical)

In [40]: wide_frame
Out[40]:
0
1

1.15. v0.10.0 (December 17, 2012)

191

pandas: powerful Python data analysis toolkit, Release 0.17.0

0
1
2
3
4

-0.681624 0.191356 1.180274

0.441522 -0.316864 -0.017062
-0.412451 -0.462580 0.422194
-0.277255 1.331263 0.585174
-1.642511 0.432560 1.218080

3
4
5
0 -0.834179 0.703043 0.166568
1 1.570114 -0.360875 -0.880096
2 0.288403 -0.487393 -0.777639
3 -0.568825 -0.719412 1.191340
4 -0.564705 -0.581790 0.286071

6
7
8
0 -0.583599 -1.201796 -1.422811
1 0.235532 0.207232 -1.983857
2 0.055865 1.383381 0.085638
3 -0.456362 0.089931 0.776079
4 0.048725 1.002440 1.276582

9
10
11
0 -0.882554 1.209871 -0.941235
1 -1.702547 -1.621234 -0.906840
2 0.246392 0.965887 0.246354
3 0.752889 -1.195795 -1.425911
4 0.054399 0.241963 -0.471786

12
13
0 0.863067 -0.336232
1 1.014601 -0.475108
2 -0.727728 -0.094414
3 -0.548829 0.774225
4 0.314510 -0.059986

14
-0.976847
-0.358944
-0.276854
0.740501
-2.069319

15
0 0.033862
1 1.262942
2 0.158399
3 1.510263
4 -1.115104
[5 rows x 16 columns]

1.15.5 Updated PyTables Support

Docs for PyTables Table format & several enhancements to the api. Here is a taste of what to expect.
In [41]: store = HDFStore('store.h5')
In [42]: df = DataFrame(randn(8, 3), index=date_range('1/1/2000', periods=8),
....:
columns=['A', 'B', 'C'])
....:
In [43]: df
Out[43]:
A
B
C
2000-01-01 -0.369325 -1.502617 -0.376280

192

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

2000-01-02 0.511936 -0.116412 -0.625256

2000-01-03 -0.550627 1.261433 -0.552429
2000-01-04 1.695803 -1.025917 -0.910942
2000-01-05 0.426805 -0.131749 0.432600
2000-01-06 0.044671 -0.341265 1.844536
2000-01-07 -2.036047 0.000830 -0.955697
2000-01-08 -0.898872 -0.725411 0.059904
[8 rows x 3 columns]
# appending data frames
In [44]: df1 = df[0:4]
In [45]: df2 = df[4:]
In [46]: store.append('df', df1)
In [47]: store.append('df', df2)
In [48]: store
Out[48]:
<class 'pandas.io.pytables.HDFStore'>
File path: store.h5
/df
frame_table (typ->appendable,nrows->8,ncols->3,indexers->[index])
# selecting the entire store
In [49]: store.select('df')
Out[49]:
A
B
2000-01-01 -0.369325 -1.502617
2000-01-02 0.511936 -0.116412
2000-01-03 -0.550627 1.261433
2000-01-04 1.695803 -1.025917
2000-01-05 0.426805 -0.131749
2000-01-06 0.044671 -0.341265
2000-01-07 -2.036047 0.000830
2000-01-08 -0.898872 -0.725411

C
-0.376280
-0.625256
-0.552429
-0.910942
0.432600
1.844536
-0.955697
0.059904

[8 rows x 3 columns]
In [50]: wp = Panel(randn(2, 5, 4), items=['Item1', 'Item2'],
....:
major_axis=date_range('1/1/2000', periods=5),
....:
minor_axis=['A', 'B', 'C', 'D'])
....:
In [51]: wp
Out[51]:
<class 'pandas.core.panel.Panel'>
Dimensions: 2 (items) x 5 (major_axis) x 4 (minor_axis)
Items axis: Item1 to Item2
Major_axis axis: 2000-01-01 00:00:00 to 2000-01-05 00:00:00
Minor_axis axis: A to D
# storing a panel
In [52]: store.append('wp',wp)
# selecting via A QUERY
In [53]: store.select('wp',

1.15. v0.10.0 (December 17, 2012)

193

pandas: powerful Python data analysis toolkit, Release 0.17.0

....:
[ Term('major_axis>20000102'), Term('minor_axis', '=', ['A','B']) ])
....:
Out[53]:
<class 'pandas.core.panel.Panel'>
Dimensions: 2 (items) x 3 (major_axis) x 2 (minor_axis)
Items axis: Item1 to Item2
Major_axis axis: 2000-01-03 00:00:00 to 2000-01-05 00:00:00
Minor_axis axis: A to B
# removing data from tables
In [54]: store.remove('wp', Term('major_axis>20000103'))
Out[54]: 8
In [55]: store.select('wp')
Out[55]:
<class 'pandas.core.panel.Panel'>
Dimensions: 2 (items) x 3 (major_axis) x 4 (minor_axis)
Items axis: Item1 to Item2
Major_axis axis: 2000-01-01 00:00:00 to 2000-01-03 00:00:00
Minor_axis axis: A to D
# deleting a store
In [56]: del store['df']
In [57]: store
Out[57]:
<class 'pandas.io.pytables.HDFStore'>
File path: store.h5
/wp
wide_table
(typ->appendable,nrows->12,ncols->2,indexers->[major_axis,minor_axis])

Enhancements
added ability to hierarchical keys
In [58]: store.put('foo/bar/bah', df)
In [59]: store.append('food/orange', df)
In [60]: store.append('food/apple',

df)

In [61]: store
Out[61]:
<class 'pandas.io.pytables.HDFStore'>
File path: store.h5
/foo/bar/bah
frame
(shape->[8,3])
/food/apple
frame_table (typ->appendable,nrows->8,ncols->3,indexers->[index])
/food/orange
frame_table (typ->appendable,nrows->8,ncols->3,indexers->[index])
/wp
wide_table
(typ->appendable,nrows->12,ncols->2,indexers->[major_ax
# remove all nodes under this level
In [62]: store.remove('food')

In [63]: store
Out[63]:
<class 'pandas.io.pytables.HDFStore'>
File path: store.h5
/foo/bar/bah
frame
(shape->[8,3])
/wp
wide_table
(typ->appendable,nrows->12,ncols->2,indexers->[major_ax

194

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

added mixed-dtype support!

In [64]: df['string'] = 'string'
In [65]: df['int']

= 1

In [66]: store.append('df',df)
In [67]: df1 = store.select('df')
In [68]: df1
Out[68]:
2000-01-01
2000-01-02
2000-01-03
2000-01-04
2000-01-05
2000-01-06
2000-01-07
2000-01-08

A
-0.369325
0.511936
-0.550627
1.695803
0.426805
0.044671
-2.036047
-0.898872

B
-1.502617
-0.116412
1.261433
-1.025917
-0.131749
-0.341265
0.000830
-0.725411

C
-0.376280
-0.625256
-0.552429
-0.910942
0.432600
1.844536
-0.955697
0.059904

string
string
string
string
string
string
string
string
string

int
1
1
1
1
1
1
1
1

[8 rows x 5 columns]
In [69]: df1.get_dtype_counts()
Out[69]:
float64
3
int64
1
object
1
dtype: int64

performance improvments on table writing

support for arbitrarily indexed dimensions
SparseSeries now has a density property (GH2384)
enable Series.str.strip/lstrip/rstrip methods to take an input argument to strip arbitrary characters (GH2411)
implement value_vars in melt to limit values to certain columns and add melt to pandas namespace
(GH2412)
Bug Fixes
added Term method of specifying where conditions (GH1996).
del store[df] now call store.remove(df) for store deletion
deleting of consecutive rows is much faster than before
min_itemsize parameter can be specified in table creation to force a minimum size for indexing columns
(the previous implementation would set the column size based on the first append)
indexing support via create_table_index (requires PyTables >= 2.3) (GH698).
appending on a store would fail if the table was not first created via put
fixed issue with missing attributes after loading a pickled dataframe (GH2431)
minor change to select and remove: require a table ONLY if where is also provided (and not None)
Compatibility

1.15. v0.10.0 (December 17, 2012)

195

pandas: powerful Python data analysis toolkit, Release 0.17.0

0.10 of HDFStore is backwards compatible for reading tables created in a prior version of pandas, however, query
terms using the prior (undocumented) methodology are unsupported. You must read in the entire file and write it out
using the new format to take advantage of the updates.

1.15.6 N Dimensional Panels (Experimental)

Adding experimental support for Panel4D and factory functions to create n-dimensional named panels. Docs for
NDim. Here is a taste of what to expect.
In [70]: p4d = Panel4D(randn(2, 2, 5, 4),
....:
labels=['Label1','Label2'],
....:
items=['Item1', 'Item2'],
....:
major_axis=date_range('1/1/2000', periods=5),
....:
minor_axis=['A', 'B', 'C', 'D'])
....:
In [71]: p4d
Out[71]:
<class 'pandas.core.panelnd.Panel4D'>
Dimensions: 2 (labels) x 2 (items) x 5 (major_axis) x 4 (minor_axis)
Labels axis: Label1 to Label2
Items axis: Item1 to Item2
Major_axis axis: 2000-01-01 00:00:00 to 2000-01-05 00:00:00
Minor_axis axis: A to D

See the full release notes or issue tracker on GitHub for a complete list.

1.16 v0.9.1 (November 14, 2012)

This is a bugfix release from 0.9.0 and includes several new features and enhancements along with a large number of
bug fixes. The new features include by-column sort order for DataFrame and Series, improved NA handling for the
rank method, masking functions for DataFrame, and intraday time-series filtering for DataFrame.

1.16.1 New features

Series.sort, DataFrame.sort, and DataFrame.sort_index can now be specified in a per-column manner to support
multiple sort orders (GH928)
In [1]: df = DataFrame(np.random.randint(0, 2, (6, 3)), columns=['A', 'B', 'C'])
In [2]: df.sort(['A', 'B'], ascending=[1, 0])
Out[2]:
A B C
0 0 1 0
2 0 0 1
1 1 1 1
5 1 1 0
3 1 0 0
4 1 0 1
[6 rows x 3 columns]

DataFrame.rank now supports additional argument values for the na_option parameter so missing values can
be assigned either the largest or the smallest rank (GH1508, GH2159)

196

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [3]: df = DataFrame(np.random.randn(6, 3), columns=['A', 'B', 'C'])

In [4]: df.ix[2:4] = np.nan
In [5]: df.rank()
Out[5]:
A
B
C
0
3
2
1
1
1
3
3
2 NaN NaN NaN
3 NaN NaN NaN
4 NaN NaN NaN
5
2
1
2
[6 rows x 3 columns]
In [6]: df.rank(na_option='top')
Out[6]:
A B C
0 6 5 4
1 4 6 6
2 2 2 2
3 2 2 2
4 2 2 2
5 5 4 5
[6 rows x 3 columns]
In [7]: df.rank(na_option='bottom')
Out[7]:
A B C
0 3 2 1
1 1 3 3
2 5 5 5
3 5 5 5
4 5 5 5
5 2 1 2
[6 rows x 3 columns]

DataFrame has new where and mask methods to select values according to a given boolean mask (GH2109,
GH2151)
DataFrame currently supports slicing via a boolean vector the same length as the DataFrame (inside
the []). The returned DataFrame has the same number of columns as the original, but is sliced on its
index.
In [8]: df = DataFrame(np.random.randn(5, 3), columns = ['A','B','C'])
In [9]: df
Out[9]:
0
1
2
3
4

A
B
C
-0.187239 -1.703664 0.613136
-0.948528 0.505346 0.017228
-2.391256 1.207381 0.853174
0.124213 -0.625597 -1.211224
-0.476548 0.649425 0.004610

[5 rows x 3 columns]

1.16. v0.9.1 (November 14, 2012)

197

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [10]: df[df['A'] > 0]

Out[10]:
A
B
C
3 0.124213 -0.625597 -1.211224
[1 rows x 3 columns]

If a DataFrame is sliced with a DataFrame based boolean condition (with the same size as the original
DataFrame), then a DataFrame the same size (index and columns) as the original is returned, with
elements that do not meet the boolean condition as NaN. This is accomplished via the new method
DataFrame.where. In addition, where takes an optional other argument for replacement.
In [11]: df[df>0]
Out[11]:
A
B
0
NaN
NaN
1
NaN 0.505346
2
NaN 1.207381
3 0.124213
NaN
4
NaN 0.649425

C
0.613136
0.017228
0.853174
NaN
0.004610

[5 rows x 3 columns]
In [12]: df.where(df>0)
Out[12]:
A
B
C
0
NaN
NaN 0.613136
1
NaN 0.505346 0.017228
2
NaN 1.207381 0.853174
3 0.124213
NaN
NaN
4
NaN 0.649425 0.004610
[5 rows x 3 columns]
In [13]: df.where(df>0,-df)
Out[13]:
A
B
C
0 0.187239 1.703664 0.613136
1 0.948528 0.505346 0.017228
2 2.391256 1.207381 0.853174
3 0.124213 0.625597 1.211224
4 0.476548 0.649425 0.004610
[5 rows x 3 columns]

Furthermore, where now aligns the input boolean condition (ndarray or DataFrame), such that partial
selection with setting is possible. This is analagous to partial setting via .ix (but on the contents rather
than the axis labels)
In [14]: df2 = df.copy()
In [15]: df2[ df2[1:4] > 0 ] = 3
In [16]: df2
Out[16]:
A
B
0 -0.187239 -1.703664
1 -0.948528 3.000000

198

C
0.613136
3.000000

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

2 -2.391256 3.000000 3.000000

3 3.000000 -0.625597 -1.211224
4 -0.476548 0.649425 0.004610
[5 rows x 3 columns]

DataFrame.mask is the inverse boolean operation of where.

In [17]: df.mask(df<=0)
Out[17]:
A
B
C
0
NaN
NaN 0.613136
1
NaN 0.505346 0.017228
2
NaN 1.207381 0.853174
3 0.124213
NaN
NaN
4
NaN 0.649425 0.004610
[5 rows x 3 columns]

Enable referencing of Excel columns by their column names (GH1936)

In [18]: xl = ExcelFile('data/test.xls')
In [19]: xl.parse('Sheet1', index_col=0, parse_dates=True,
....:
parse_cols='A:D')
....:
Out[19]:
A
B
C
2000-01-03 0.980269 3.685731 -0.364217
2000-01-04 1.047916 -0.041232 -0.161812
2000-01-05 0.498581 0.731168 -0.537677
2000-01-06 1.120202 1.567621 0.003641
2000-01-07 -0.487094 0.571455 -1.611639
2000-01-10 0.836649 0.246462 0.588543
2000-01-11 -0.157161 1.340307 1.195778
[7 rows x 3 columns]

Added option to disable pandas-style tick locators and formatters using series.plot(x_compat=True) or pandas.plot_params[x_compat] = True (GH2205)
Existing TimeSeries methods at_time and between_time were added to DataFrame (GH2149)
DataFrame.dot can now accept ndarrays (GH2042)
DataFrame.drop now supports non-unique indexes (GH2101)
Panel.shift now supports negative periods (GH2164)
DataFrame now support unary ~ operator (GH2110)

1.16.2 API changes

Upsampling data with a PeriodIndex will result in a higher frequency TimeSeries that spans the original time
window
In [20]: prng = period_range('2012Q1', periods=2, freq='Q')
In [21]: s = Series(np.random.randn(len(prng)), prng)

1.16. v0.9.1 (November 14, 2012)

199

pandas: powerful Python data analysis toolkit, Release 0.17.0

In [22]: s.resample('M')
Out[22]:
2012-01
-0.508759
2012-02
NaN
2012-03
NaN
2012-04
-0.599515
2012-05
NaN
2012-06
NaN
Freq: M, dtype: float64

Period.end_time now returns the last nanosecond in the time interval (GH2124, GH2125, GH1764)
In [23]: p = Period('2012')
In [24]: p.end_time
Out[24]: Timestamp('2012-12-31 23:59:59.999999999')

File parsers no longer coerce to float or bool for columns that have custom converters specified (GH2184)
In [25]: data = 'A,B,C\n00001,001,5\n00002,002,6'
In [26]: read_csv(StringIO(data), converters={'A' : lambda x: x.strip()})
Out[26]:
A B C
0 00001 1 5
1 00002 2 6
[2 rows x 3 columns]

See the full release notes or issue tracker on GitHub for a complete list.

1.17 v0.9.0 (October 7, 2012)

This is a major release from 0.8.1 and includes several new features and enhancements along with a large number of
bug fixes. New features include vectorized unicode encoding/decoding for Series.str, to_latex method to DataFrame,
more flexible parsing of boolean values, and enabling the download of options data from Yahoo! Finance.

1.17.1 New features

Add encode and decode for unicode handling to vectorized string processing methods in Series.str (GH1706)
Add DataFrame.to_latex method (GH1735)
Add convenient expanding window equivalents of all rolling_* ops (GH1785)
Add Options class to pandas.io.data for fetching options data from Yahoo! Finance (GH1748, GH1739)
More flexible parsing of boolean values (Yes, No, TRUE, FALSE, etc) (GH1691, GH1295)
Add level parameter to Series.reset_index
TimeSeries.between_time can now select times across midnight (GH1871)
Series constructor can now handle generator as input (GH1679)
DataFrame.dropna can now take multiple axes (tuple/list) as input (GH924)
Enable skip_footer parameter in ExcelFile.parse (GH1843)

200

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.17.2 API changes

The default column names when header=None and no columns names passed to functions like read_csv
has changed to be more Pythonic and amenable to attribute access:
In [1]: data = '0,0,1\n1,1,0\n0,1,0'
In [2]: df = read_csv(StringIO(data), header=None)
In [3]: df
Out[3]:
0 1 2
0 0 0 1
1 1 1 0
2 0 1 0
[3 rows x 3 columns]

Creating a Series from another Series, passing an index, will cause reindexing to happen inside rather than treating the Series like an ndarray. Technically improper usages like Series(df[col1], index=df[col2])
that worked before by accident (this was never intended) will lead to all NA Series in some cases. To be perfectly clear:
In [4]: s1 = Series([1, 2, 3])
In [5]: s1
Out[5]:
0
1
1
2
2
3
dtype: int64
In [6]: s2 = Series(s1, index=['foo', 'bar', 'baz'])
In [7]: s2
Out[7]:
foo
NaN
bar
NaN
baz
NaN
dtype: float64

Deprecated day_of_year API removed from PeriodIndex, use dayofyear (GH1723)

Dont modify NumPy suppress printoption to True at import time
The internal HDF5 data arrangement for DataFrames has been transposed. Legacy files will still be readable by
HDFStore (GH1834, GH1824)
Legacy cruft removed: pandas.stats.misc.quantileTS
Use ISO8601 format for Period repr: monthly, daily, and on down (GH1776)
Empty DataFrame columns are now created as object dtype. This will prevent a class of TypeErrors that was
occurring in code where the dtype of a column would depend on the presence of data or not (e.g. a SQL query
having results) (GH1783)
Setting parts of DataFrame/Panel using ix now aligns input Series/DataFrame (GH1630)
first and last methods in GroupBy no longer drop non-numeric columns (GH1809)

1.17. v0.9.0 (October 7, 2012)

201

pandas: powerful Python data analysis toolkit, Release 0.17.0

Resolved inconsistencies in specifying custom NA values in text parser. na_values of type dict no longer
override default NAs unless keep_default_na is set to false explicitly (GH1657)
DataFrame.dot will not do data alignment, and also work with Series (GH1915)
See the full release notes or issue tracker on GitHub for a complete list.

1.18 v0.8.1 (July 22, 2012)

This release includes a few new features, performance enhancements, and over 30 bug fixes from 0.8.0. New features
include notably NA friendly string processing functionality and a series of new plot types and options.

1.18.1 New features

Add vectorized string processing methods accessible via Series.str (GH620)
Add option to disable adjustment in EWMA (GH1584)
Radviz plot (GH1566)
Parallel coordinates plot
Bootstrap plot
Per column styles and secondary y-axis plotting (GH1559)
New datetime converters millisecond plotting (GH1599)
Add option to disable sparse display of hierarchical indexes (GH1538)
Series/DataFrames set_index method can append levels to an existing Index/MultiIndex (GH1569,
GH1577)

1.18.2 Performance improvements

Improved implementation of rolling min and max (thanks to Bottleneck !)
Add accelerated median GroupBy option (GH1358)
Significantly improve the performance of parsing ISO8601-format date strings with DatetimeIndex or
to_datetime (GH1571)
Improve the performance of GroupBy on single-key aggregations and use with Categorical types
Significant datetime parsing performance improvments

1.19 v0.8.0 (June 29, 2012)

This is a major release from 0.7.3 and includes extensive work on the time series handling and processing infrastructure
as well as a great deal of new functionality throughout the library. It includes over 700 commits from more than 20
distinct authors. Most pandas 0.7.3 and earlier users should not experience any issues upgrading, but due to the
migration to the NumPy datetime64 dtype, there may be a number of bugs and incompatibilities lurking. Lingering
incompatibilities will be fixed ASAP in a 0.8.1 release if necessary. See the full release notes or issue tracker on
GitHub for a complete list.

202

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.19.1 Support for non-unique indexes

All objects can now work with non-unique indexes. Data alignment / join operations work according to SQL join
semantics (including, if application, index duplication in many-to-many joins)

1.19.2 NumPy datetime64 dtype and 1.6 dependency

Time series data are now represented using NumPys datetime64 dtype; thus, pandas 0.8.0 now requires at least NumPy
1.6. It has been tested and verified to work with the development version (1.7+) of NumPy as well which includes some
significant user-facing API changes. NumPy 1.6 also has a number of bugs having to do with nanosecond resolution
data, so I recommend that you steer clear of NumPy 1.6s datetime64 API functions (though limited as they are) and
only interact with this data using the interface that pandas provides.
See the end of the 0.8.0 section for a porting guide listing potential issues for users migrating legacy codebases from
pandas 0.7 or earlier to 0.8.0.
Bug fixes to the 0.7.x series for legacy NumPy < 1.6 users will be provided as they arise. There will be no more further
development in 0.7.x beyond bug fixes.

1.19.3 Time series changes and improvements

Note: With this release, legacy scikits.timeseries users should be able to port their code to use pandas.
Note: See documentation for overview of pandas timeseries API.
New datetime64 representation speeds up join operations and data alignment, reduces memory usage, and
improve serialization / deserialization performance significantly over datetime.datetime
High performance and flexible resample method for converting from high-to-low and low-to-high frequency.
Supports interpolation, user-defined aggregation functions, and control over how the intervals and result labeling
are defined. A suite of high performance Cython/C-based resampling functions (including Open-High-LowClose) have also been implemented.
Revamp of frequency aliases and support for frequency shortcuts like 15min, or 1h30min
New DatetimeIndex class supports both fixed frequency and irregular time series. Replaces now deprecated
DateRange class
New PeriodIndex and Period classes for representing time spans and performing calendar logic, including the 12 fiscal quarterly frequencies <timeseries.quarterly>. This is a partial port of, and a substantial
enhancement to, elements of the scikits.timeseries codebase. Support for conversion between PeriodIndex and
DatetimeIndex
New Timestamp data type subclasses datetime.datetime, providing the same interface while enabling working
with nanosecond-resolution data. Also provides easy time zone conversions.
Enhanced support for time zones. Add tz_convert and tz_lcoalize methods to TimeSeries and DataFrame.
All timestamps are stored as UTC; Timestamps from DatetimeIndex objects with time zone set will be localized
to localtime. Time zone conversions are therefore essentially free. User needs to know very little about pytz
library now; only time zone names as as strings are required. Time zone-aware timestamps are equal if and only
if their UTC timestamps match. Operations between time zone-aware time series with different time zones will
result in a UTC-indexed time series.
Time series string indexing conveniences / shortcuts: slice years, year and month, and index values with strings
Enhanced time series plotting; adaptation of scikits.timeseries matplotlib-based plotting code
1.19. v0.8.0 (June 29, 2012)

203

pandas: powerful Python data analysis toolkit, Release 0.17.0

New date_range, bdate_range, and period_range factory functions

Robust frequency inference function infer_freq and inferred_freq property of DatetimeIndex, with option
to infer frequency on construction of DatetimeIndex
to_datetime function efficiently parses array of strings to DatetimeIndex. DatetimeIndex will parse array or
list of strings to datetime64
Optimized support for datetime64-dtype data in Series and DataFrame columns
New NaT (Not-a-Time) type to represent NA in timestamp arrays
Optimize Series.asof for looking up as of values for arrays of timestamps
Milli, Micro, Nano date offset objects
Can index time series with datetime.time objects to select all data at particular time of day
(TimeSeries.at_time) or between two times (TimeSeries.between_time)
Add tshift method for leading/lagging using the frequency (if any) of the index, as opposed to a naive lead/lag
using shift

1.19.4 Other new features

New cut and qcut functions (like Rs cut function) for computing a categorical variable from a continuous
variable by binning values either into value-based (cut) or quantile-based (qcut) bins
Rename Factor to Categorical and add a number of usability features
Add limit argument to fillna/reindex
More flexible multiple function application in GroupBy, and can pass list (name, function) tuples to get result in
particular order with given names
Add flexible replace method for efficiently substituting values
Enhanced read_csv/read_table for reading time series data and converting multiple columns to dates
Add comments option to parser functions: read_csv, etc.
Add :refdayfirst <io.dayfirst> option to parser functions for parsing international DD/MM/YYYY dates
Allow the user to specify the CSV reader dialect to control quoting etc.
Handling thousands separators in read_csv to improve integer parsing.
Enable unstacking of multiple levels in one shot. Alleviate pivot_table bugs (empty columns being introduced)
Move to klib-based hash tables for indexing; better performance and less memory usage than Pythons dict
Add first, last, min, max, and prod optimized GroupBy functions
New ordered_merge function
Add flexible comparison instance methods eq, ne, lt, gt, etc. to DataFrame, Series
Improve scatter_matrix plotting function and add histogram or kernel density estimates to diagonal
Add kde plot option for density plots
Support for converting DataFrame to R data.frame through rpy2
Improved support for complex numbers in Series and DataFrame
Add pct_change method to all data structures

204

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Add max_colwidth configuration option for DataFrame console output

Interpolate Series values using index values
Can select multiple columns from GroupBy
Add update methods to Series/DataFrame for updating values in place
Add any and all method to DataFrame

1.19.5 New plotting methods

Series.plot now supports a secondary_y option:
In [1]: plt.figure()
Out[1]: <matplotlib.figure.Figure at 0xa852f34c>
In [2]: fx['FR'].plot(style='g')
Out[2]: <matplotlib.axes._subplots.AxesSubplot at 0xa852f9ac>
In [3]: fx['IT'].plot(style='k--', secondary_y=True)
Out[3]: <matplotlib.axes._subplots.AxesSubplot at 0xa8d702cc>

../_static/whatsnew_secondary_y.png

Vytautas Jancauskas, the 2012 GSOC participant, has added many new plot types. For example, kde is a new
option:
In [4]: s = Series(np.concatenate((np.random.randn(1000),
...:
np.random.randn(1000) * 0.5 + 3)))
...:
In [5]: plt.figure()
Out[5]: <matplotlib.figure.Figure at 0xaf3dd2ac>
In [6]: s.hist(normed=True, alpha=0.2)
Out[6]: <matplotlib.axes._subplots.AxesSubplot at 0xa862da2c>
In [7]: s.plot(kind='kde')
Out[7]: <matplotlib.axes._subplots.AxesSubplot at 0xa862da2c>

../_static/whatsnew_kde.png

See the plotting page for much more.

1.19. v0.8.0 (June 29, 2012)

205

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.19.6 Other API changes

Deprecation of offset, time_rule, and timeRule arguments names in time series functions. Warnings
will be printed until pandas 0.9 or 1.0.

1.19.7 Potential porting issues for pandas <= 0.7.3 users

The major change that may affect you in pandas 0.8.0 is that time series indexes use NumPys datetime64 data
type instead of dtype=object arrays of Pythons built-in datetime.datetime objects. DateRange has been
replaced by DatetimeIndex but otherwise behaved identically. But, if you have code that converts DateRange
or Index objects that used to contain datetime.datetime values to plain NumPy arrays, you may have bugs
lurking with code using scalar values because you are handing control over to NumPy:
In [8]: import datetime
In [9]: rng = date_range('1/1/2000', periods=10)
In [10]: rng[5]
Out[10]: Timestamp('2000-01-06 00:00:00', offset='D')
In [11]: isinstance(rng[5], datetime.datetime)
Out[11]: True
In [12]: rng_asarray = np.asarray(rng)
In [13]: scalar_val = rng_asarray[5]
In [14]: type(scalar_val)
Out[14]: numpy.datetime64

pandass Timestamp object is a subclass of datetime.datetime that has nanosecond support (the
nanosecond field store the nanosecond value between 0 and 999). It should substitute directly into any code that
used datetime.datetime values before. Thus, I recommend not casting DatetimeIndex to regular NumPy
arrays.
If you have code that requires an array of datetime.datetime objects, you have a couple of options. First, the
asobject property of DatetimeIndex produces an array of Timestamp objects:
In [15]: stamp_array = rng.asobject
In [16]: stamp_array
Out[16]:
Index([2000-01-01 00:00:00, 2000-01-02 00:00:00, 2000-01-03 00:00:00,
2000-01-04 00:00:00, 2000-01-05 00:00:00, 2000-01-06 00:00:00,
2000-01-07 00:00:00, 2000-01-08 00:00:00, 2000-01-09 00:00:00,
2000-01-10 00:00:00],
dtype='object')
In [17]: stamp_array[5]
Out[17]: Timestamp('2000-01-06 00:00:00', offset='D')

To get an array of proper datetime.datetime objects, use the to_pydatetime method:

In [18]: dt_array = rng.to_pydatetime()
In [19]: dt_array
Out[19]:

206

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

array([datetime.datetime(2000,
datetime.datetime(2000,
datetime.datetime(2000,
datetime.datetime(2000,
datetime.datetime(2000,
datetime.datetime(2000,
datetime.datetime(2000,
datetime.datetime(2000,
datetime.datetime(2000,
datetime.datetime(2000,

1,
1,
1,
1,
1,
1,
1,
1,
1,
1,

1, 0, 0),
2, 0, 0),
3, 0, 0),
4, 0, 0),
5, 0, 0),
6, 0, 0),
7, 0, 0),
8, 0, 0),
9, 0, 0),
10, 0, 0)], dtype=object)

In [20]: dt_array[5]
Out[20]: datetime.datetime(2000, 1, 6, 0, 0)

matplotlib knows how to handle datetime.datetime but not Timestamp objects. While I recommend that you
plot time series using TimeSeries.plot, you can either use to_pydatetime or register a converter for the
Timestamp type. See matplotlib documentation for more on this.
Warning: There are bugs in the user-facing API with the nanosecond datetime64 unit in NumPy 1.6. In particular,
the string version of the array shows garbage values, and conversion to dtype=object is similarly broken.
In [21]: rng = date_range('1/1/2000', periods=10)
In [22]: rng
Out[22]:
DatetimeIndex(['2000-01-01', '2000-01-02', '2000-01-03', '2000-01-04',
'2000-01-05', '2000-01-06', '2000-01-07', '2000-01-08',
'2000-01-09', '2000-01-10'],
dtype='datetime64[ns]', freq='D')
In [23]: np.asarray(rng)
Out[23]:
array(['2000-01-01T01:00:00.000000000+0100',
'2000-01-02T01:00:00.000000000+0100',
'2000-01-03T01:00:00.000000000+0100',
'2000-01-04T01:00:00.000000000+0100',
'2000-01-05T01:00:00.000000000+0100',
'2000-01-06T01:00:00.000000000+0100',
'2000-01-07T01:00:00.000000000+0100',
'2000-01-08T01:00:00.000000000+0100',
'2000-01-09T01:00:00.000000000+0100',
'2000-01-10T01:00:00.000000000+0100'], dtype='datetime64[ns]')
In [24]: converted = np.asarray(rng, dtype=object)
In [25]: converted[5]
Out[25]: 947116800000000000L

Trust me: dont panic. If you are using NumPy 1.6 and restrict your interaction with datetime64 values to
pandass API you will be just fine. There is nothing wrong with the data-type (a 64-bit integer internally); all of the
important data processing happens in pandas and is heavily tested. I strongly recommend that you do not work
directly with datetime64 arrays in NumPy 1.6 and only use the pandas API.
Support for non-unique indexes: In the latter case, you may have code inside a try:... catch: block that
failed due to the index not being unique. In many cases it will no longer fail (some method like append still check for
uniqueness unless disabled). However, all is not lost: you can inspect index.is_unique and raise an exception
explicitly if it is False or go to a different code branch.

1.19. v0.8.0 (June 29, 2012)

207

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.20 v.0.7.3 (April 12, 2012)

This is a minor release from 0.7.2 and fixes many minor bugs and adds a number of nice new features. There are
also a couple of API changes to note; these should not affect very many users, and we are inclined to call them bug
fixes even though they do constitute a change in behavior. See the full release notes or issue tracker on GitHub for a
complete list.

1.20.1 New features

New fixed width file reader, read_fwf
New scatter_matrix function for making a scatter plot matrix
from pandas.tools.plotting import scatter_matrix
scatter_matrix(df, alpha=0.2)

Add stacked argument to Series and DataFrames plot method for stacked bar plots.
df.plot(kind='bar', stacked=True)

208

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

df.plot(kind='barh', stacked=True)

Add log x and y scaling options to DataFrame.plot and Series.plot

Add kurt methods to Series and DataFrame for computing kurtosis

1.20.2 NA Boolean Comparison API Change

Reverted some changes to how NA values (represented typically as NaN or None) are handled in non-numeric Series:
In [1]: series = Series(['Steve', np.nan, 'Joe'])
In [2]: series == 'Steve'
Out[2]:
0
True

1.20. v.0.7.3 (April 12, 2012)

209

pandas: powerful Python data analysis toolkit, Release 0.17.0

1
False
2
False
dtype: bool
In [3]: series != 'Steve'
Out[3]:
0
False
1
True
2
True
dtype: bool

In comparisons, NA / NaN will always come through as False except with != which is True. Be very careful with
boolean arithmetic, especially negation, in the presence of NA data. You may wish to add an explicit NA filter into
boolean array operations if you are worried about this:
In [4]: mask = series == 'Steve'
In [5]: series[mask & series.notnull()]
Out[5]:
0
Steve
dtype: object

While propagating NA in comparisons may seem like the right behavior to some users (and you could argue on purely
technical grounds that this is the right thing to do), the evaluation was made that propagating NA everywhere, including
in numerical arrays, would cause a large amount of problems for users. Thus, a practicality beats purity approach
was taken. This issue may be revisited at some point in the future.

1.20.3 Other API Changes

When calling apply on a grouped Series, the return value will also be a Series, to be more consistent with the
groupby behavior with DataFrame:
In [6]: df = DataFrame({'A' : ['foo', 'bar', 'foo', 'bar',
...:
'foo', 'bar', 'foo', 'foo'],
...:
'B' : ['one', 'one', 'two', 'three',
...:
'two', 'two', 'one', 'three'],
...:
'C' : np.random.randn(8), 'D' : np.random.randn(8)})
...:
In [7]: df
Out[7]:
A
B
0 foo
one
1 bar
one
2 foo
two
3 bar three
4 foo
two
5 bar
two
6 foo
one
7 foo three

C
D
-0.032419 0.237288
-1.581534 0.514679
-0.912061 -1.488101
0.209500 1.018514
-0.675890 -1.488840
0.055228 -1.355434
-1.079181 0.661979
-1.631882 0.708958

[8 rows x 4 columns]
In [8]: grouped = df.groupby('A')['C']
In [9]: grouped.describe()
Out[9]:

210

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

A
bar

count
mean
std
min
25%
50%
75%

3.000000
-0.438936
0.992521
-1.581534
-0.763153
0.055228
0.132364
...
foo mean
-0.866287
std
0.584196
min
-1.631882
25%
-1.079181
50%
-0.912061
75%
-0.675890
max
-0.032419
dtype: float64
In [10]: grouped.apply(lambda x: x.order()[-2:]) # top 2 values
Out[10]:
A
bar 5
0.055228
3
0.209500
foo 4
-0.675890
0
-0.032419
dtype: float64

1.21 v.0.7.2 (March 16, 2012)

This release targets bugs in 0.7.1, and adds a few minor features.

1.21.1 New features

Add additional tie-breaking methods in DataFrame.rank (GH874)
Add ascending parameter to rank in Series, DataFrame (GH875)
Add coerce_float option to DataFrame.from_records (GH893)
Add sort_columns parameter to allow unsorted plots (GH918)
Enable column access via attributes on GroupBy (GH882)
Can pass dict of values to DataFrame.fillna (GH661)
Can select multiple hierarchical groups by passing list of values in .ix (GH134)
Add axis option to DataFrame.fillna (GH174)
Add level keyword to drop for dropping values from a level (GH159)

1.21.2 Performance improvements

Use khash for Series.value_counts, add raw function to algorithms.py (GH861)
Intercept __builtin__.sum in groupby (GH885)

1.21. v.0.7.2 (March 16, 2012)

211

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.22 v.0.7.1 (February 29, 2012)

This release includes a few new features and addresses over a dozen bugs in 0.7.0.

1.22.1 New features

Add to_clipboard function to pandas namespace for writing objects to the system clipboard (GH774)
Add itertuples method to DataFrame for iterating through the rows of a dataframe as tuples (GH818)
Add ability to pass fill_value and method to DataFrame and Series align method (GH806, GH807)
Add fill_value option to reindex, align methods (GH784)
Enable concat to produce DataFrame from Series (GH787)
Add between method to Series (GH802)
Add HTML representation hook to DataFrame for the IPython HTML notebook (GH773)
Support for reading Excel 2007 XML documents using openpyxl

1.22.2 Performance improvements

Improve performance and memory usage of fillna on DataFrame
Can concatenate a list of Series along axis=1 to obtain a DataFrame (GH787)

1.23 v.0.7.0 (February 9, 2012)

1.23.1 New features
New unified merge function for efficiently performing full gamut of database / relational-algebra operations.
Refactored existing join methods to use the new infrastructure, resulting in substantial performance gains
(GH220, GH249, GH267)
New unified concatenation function for concatenating Series, DataFrame or Panel objects along an axis.
Can form union or intersection of the other axes. Improves performance of Series.append and
DataFrame.append (GH468, GH479, GH273)
Can pass multiple DataFrames to DataFrame.append to concatenate (stack) and multiple Series to
Series.append too
Can pass list of dicts (e.g., a list of JSON objects) to DataFrame constructor (GH526)
You can now set multiple columns in a DataFrame via __getitem__, useful for transformation (GH342)
Handle differently-indexed output values in DataFrame.apply (GH498)
In [1]: df = DataFrame(randn(10, 4))
In [2]: df.apply(lambda x: x.describe())
Out[2]:
0
1
2
3
count 10.000000 10.000000 10.000000 10.000000
mean
0.058434
0.008207
0.449898 -0.064109
std
0.959629
1.126010
0.784723
0.650405

212

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

min
25%
50%
75%
max

-1.819334
-0.365166
0.116057
0.616168
1.387310

-1.607906
-0.973095
0.179112
0.807868
1.521442

-1.275249
0.100811
0.718606
0.851809
1.437656

-1.200953
-0.369865
-0.057095
0.237069
1.051356

[8 rows x 4 columns]

Add reorder_levels method to Series and DataFrame (GH534)

Add dict-like get function to DataFrame and Panel (GH521)
Add DataFrame.iterrows method for efficiently iterating through the rows of a DataFrame
Add DataFrame.to_panel with code adapted from LongPanel.to_long
Add reindex_axis method added to DataFrame
Add level option to binary arithmetic functions on DataFrame and Series
Add level option to the reindex and align methods on Series and DataFrame for broadcasting values
across a level (GH542, GH552, others)
Add attribute-based item access to Panel and add IPython completion (GH563)
Add logy option to Series.plot for log-scaling on the Y axis
Add index and header options to DataFrame.to_string
Can pass multiple DataFrames to DataFrame.join to join on index (GH115)
Can pass multiple Panels to Panel.join (GH115)
Added justify argument to DataFrame.to_string to allow different alignment of column headers
Add sort option to GroupBy to allow disabling sorting of the group keys for potential speedups (GH595)
Can pass MaskedArray to Series constructor (GH563)
Add Panel item access via attributes and IPython completion (GH554)
Implement DataFrame.lookup, fancy-indexing analogue for retrieving values given a sequence of row and
column labels (GH338)
Can pass a list of functions to aggregate with groupby on a DataFrame, yielding an aggregated result with
hierarchical columns (GH166)
Can call cummin and cummax on Series and DataFrame to get cumulative minimum and maximum, respectively (GH647)
value_range added as utility function to get min and max of a dataframe (GH288)
Added encoding argument to read_csv, read_table, to_csv and from_csv for non-ascii text
(GH717)
Added abs method to pandas objects
Added crosstab function for easily computing frequency tables
Added isin method to index objects
Added level argument to xs method of DataFrame.

1.23. v.0.7.0 (February 9, 2012)

213

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.23.2 API Changes to integer indexing

One of the potentially riskiest API changes in 0.7.0, but also one of the most important, was a complete review of how
integer indexes are handled with regard to label-based indexing. Here is an example:
In [3]: s = Series(randn(10), index=range(0, 20, 2))
In [4]: s
Out[4]:
0
0.041867
2
1.503116
4
-0.841265
6
-1.578003
8
-0.273728
10
1.755240
12
-0.705788
14
-0.351950
16
1.507974
18
0.419219
dtype: float64
In [5]: s[0]
Out[5]: 0.041867372914288915
In [6]: s[2]
Out[6]: 1.5031163945003796
In [7]: s[4]
Out[7]: -0.84126511615728994

This is all exactly identical to the behavior before. However, if you ask for a key not contained in the Series, in
versions 0.6.1 and prior, Series would fall back on a location-based lookup. This now raises a KeyError:
In [2]: s[1]
KeyError: 1

This change also has the same impact on DataFrame:

In [3]: df = DataFrame(randn(8, 4), index=range(0, 16, 2))
In [4]: df
0
0
0.88427
2
0.14451
4 -1.44779
6 -0.26598
8 -0.58776
10 0.10940
12 -1.16919
14 -0.07337

1
0.3363
-0.1415
-0.9186
-2.4184
0.3144
-0.7175
-0.3087
0.3410

2
3
-0.1787 0.03162
0.2504 0.58374
-1.4996 0.27163
-0.2658 0.11503
-0.8566 0.61941
-1.0108 0.47990
-0.6049 -0.43544
0.0424 -0.16037

In [5]: df.ix[3]
KeyError: 3

In order to support purely integer-based indexing, the following methods have been added:

214

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

Method
Series.iget_value(i)
Series.iget(i)
DataFrame.irow(i)
DataFrame.icol(j)
DataFrame.iget_value(i, j)

Description
Retrieve value stored at location i
Alias for iget_value
Retrieve the i-th row
Retrieve the j-th column
Retrieve the value at row i and column j

1.23.3 API tweaks regarding label-based slicing

Label-based slicing using ix now requires that the index be sorted (monotonic) unless both the start and endpoint are
contained in the index:
In [8]: s = Series(randn(6), index=list('gmkaec'))
In [9]: s
Out[9]:
g
0.647633
m
-0.147670
k
-0.759803
a
-0.757308
e
-1.921164
c
-1.093529
dtype: float64

Then this is OK:

In [10]: s.ix['k':'e']
Out[10]:
k
-0.759803
a
-0.757308
e
-1.921164
dtype: float64

But this is not:

In [12]: s.ix['b':'h']
KeyError 'b'

If the index had been sorted, the range selection would have been possible:
In [11]: s2 = s.sort_index()
In [12]: s2
Out[12]:
a
-0.757308
c
-1.093529
e
-1.921164
g
0.647633
k
-0.759803
m
-0.147670
dtype: float64
In [13]: s2.ix['b':'h']
Out[13]:
c
-1.093529
e
-1.921164
g
0.647633
dtype: float64

1.23. v.0.7.0 (February 9, 2012)

215

pandas: powerful Python data analysis toolkit, Release 0.17.0

1.23.4 Changes to Series [] operator

As as notational convenience, you can pass a sequence of labels or a label slice to a Series when getting and setting
values via [] (i.e. the __getitem__ and __setitem__ methods). The behavior will be the same as passing
similar input to ix except in the case of integer indexing:
In [14]: s = Series(randn(6), index=list('acegkm'))
In [15]: s
Out[15]:
a
-0.592157
c
-0.715074
e
-0.616193
g
-0.335468
k
-0.392051
m
-0.189537
dtype: float64
In [16]: s[['m', 'a', 'c', 'e']]
Out[16]:
m
-0.189537
a
-0.592157
c
-0.715074
e
-0.616193
dtype: float64
In [17]: s['b':'l']
Out[17]:
c
-0.715074
e
-0.616193
g
-0.335468
k
-0.392051
dtype: float64
In [18]: s['c':'k']
Out[18]:
c
-0.715074
e
-0.616193
g
-0.335468
k
-0.392051
dtype: float64

In the case of integer indexes, the behavior will be exactly as before (shadowing ndarray):
In [19]: s = Series(randn(6), index=range(0, 12, 2))
In [20]: s[[4, 0, 2]]
Out[20]:
4
0.319635
0
0.886170
2
-1.125894
dtype: float64
In [21]: s[1:5]
Out[21]:
2
-1.125894
4
0.319635
6
0.998222
8
0.091743

216

Chapter 1. Whats New

pandas: powerful Python data analysis toolkit, Release 0.17.0

dtype: float64

If you wish to do indexing with sequences and slicing on an integer index with label semantics, use ix.

1.23.5 Other API Changes

The deprecated LongPanel class has been completely removed
If Series.sort is called on a column of a DataFrame, an exception will now be raised. Before it was
possible to accidentally mutate a DataFrames column by doing df[col].sort() instead of the side-effect
free method df[col].order() (GH316)
Miscellaneous renames and deprecations which will (harmlessly) raise FutureWarning
drop added as an optional parameter to DataFrame.reset_index (GH699)

1.23.6 Performance improvements

Cythonized GroupBy aggregations no longer presort the data, thus achieving a significant speedup (GH93).
GroupBy aggregations with Python functions significantly sped up by clever manipulation of the ndarray data
type in Cython (GH496).
Better error message in DataFrame constructor when passed column labels dont match data (GH497)
Substantially improve performance of multi-GroupBy aggregation when a Python function is passed, reuse
ndarray object in Cython (GH496)
Can store objects indexed by tuples and floats in HDFStore (GH492)
Dont print length by default in Series.to_string, add length option (GH489)
Improve Cython code for multi-groupby to aggregate without having to sort the data (GH93)
Improve MultiIndex reindexing speed by storing tuples in the MultiIndex, test for backwards unpickling compatibility
Improve column reindexing performance by using specialized Cython take function
Further performance tweaking of Series.__getitem__ for standard use cases
Avoid Index dict creation in some cases (i.e. when getting slices, etc.), regression from prior versions
Friendlier error message in setup.py if NumPy not installed
Use common set of NA-handling operations (sum, mean, etc.) in Panel class also (GH536)
Default name assignment when calling reset_index on DataFrame with a regular (non-hierarchical) index
(GH476)
Use Cythonized groupers when possible in Series/DataFrame stat ops with level parameter passed (GH545)
Ported skiplist data structure to C to speed up rolling_median by about 5-10x in most typical use cases
(GH374)

1.24 v.0.6.1 (December 13, 2011)

1.24.1 New features
Can append single rows (as Series) to a DataFrame
1.24. v.0.6.1 (December 13, 2011)

217

pandas: powerful Python data analysis toolkit, Release 0.17.0

Add Spearman and Kendall rank correlation options to Series.corr and DataFrame.corr (GH428)
Added get_value and set_value methods to Series, DataFrame, and Panel for very low-overhead access
(>2x faster in many cases) to scalar elements (GH437, GH438). set_value is capable of producing an
enlarged object.
Add PyQt table widget to sandbox (GH435)
DataFrame.align can accept Series arguments and an axis option (GH461)
Implement new SparseArray and SparseList data structures. SparseSeries now derives from SparseArray
(GH463)
Better console printing options (GH453)
Implement fast data ranking for Series and DataFrame, fast versions of scipy.stats.rankdata (GH428)
Implement DataFrame.from_items alternate constructor (GH444)
DataFrame.convert_objects method for inferring better dt