Coverage decreased (-0.2%) to 81.25% when pulling ee6abc9 on pandas_notebook_fix into bb53160 on master.

Current coverage is 79.92% (diff: 15.15%)

Merging #192 into master will decrease coverage by 1.47%

@@             master       #192   diff @@
==========================================
  Files             8          8          
  Lines           527        538    +11   
  Methods           0          0          
  Messages          0          0          
  Branches         99        101     +2   
==========================================
+ Hits            429        430     +1   
- Misses           98        108    +10   
  Partials          0          0          

Powered by Codecov. Last update 1b4d483...7f6c45c

Hmm I'd asked about doing it this way instead when creating this feature but since there didn't seem to be a clear advantage back then we stuck witht the current version... this PR is not backwards compatible, but could be fixed.

Yes indeed I thought about it afterwards, we can simply detect whether the
supplied bar argument is a class or an object, and then act accordingly,
this will be backward compatible then :)
Le 5 Jul. 2016 09:07, "Casper da Costa-Luis" [email protected] a
écrit :

Hmm I'd asked about doing it this way instead when creating this feature
but since there didn't seem to be a clear advantage back then we stuck
witht the current version... this PR is not backwards compatible, but could
be fixed.

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#192 (comment), or mute
the thread
https://github.com/notifications/unsubscribe/ABES3m2JKDZPoAs2WpKLN6npbuD0zWZXks5qSgKxgaJpZM4JD81F
.

I added a deprecation message using tqdm.write(), because I didn't like the warnings module for our purpose since it suppresses warnings by default now as discussed here.

Coverage decreased (-0.8%) to 80.639% when pulling 2e74274 on pandas_notebook_fix into bb53160 on master.

Coverage decreased (-0.8%) to 80.639% when pulling 2e74274 on pandas_notebook_fix into bb53160 on master.

Coverage decreased (-0.8%) to 80.639% when pulling 787f3b2 on pandas_notebook_fix into ce2eace on master.

Wow that's great @casperdcl , thank's a lot!

Just out of curiosity, why did you make tqdm_pandas a classmethod? What is the advantage in this case?

Coverage decreased (-3.6%) to 77.758% when pulling f020419 on pandas_notebook_fix into ce2eace on master.

because the way you'd re-written it in this PR was function(Class, ...) which is precisely what class methods are for! The nice thing about this is all sub classes automatically inherit the class method, with the correct child class passed as the first argument when invoked.

combined with the fact that an instance is created within the inner function, this means .pandas() only needs to be called once in a program (more precisely, once every time you want to use a different tqdm-like class to handle progress_apply). as in:

tqdm.pandas()
... # any number of `progress_apply` calls will go to terminal
tqdm_gui.pandas()
... # now use gui instead

re: python warning confusions, much sad such is. was vaguely aware of this; surely there's a better solution. We could always class TqdmDeprecationWarning(SomethingAppropriate)?

@casperdcl Great about the classmethod, talk about the low entry high ceiling of python, I still often learn new stuff XD

I thought about doing a class for the warning, but how would you raise it? If you raise it as an exception, then the old way of using tqdm_pandas will need to be specifically caught by users, so it breaks retrocompatibility. I do not know of any other way to raise a non-blocking warning in Python than just printing...

But yeah it would be nice to have a standardized way of raising deprecation warnings, because I guess we will get more and more as time goes.

@lrq3000 There is! Here's a doc link! ;)

It's just warnings.DeprecationWarning. Learned some new stuff - again!

You can raise warnings using warnings.warn(). Does not crash program.

@CrazyPython Yes but the warning module doesn't emit any output to sys.stderr by default now as reported here except if you modify the filters at the module level, which is considered a bad practice. I tried it with Python 2.7 just to make sure and indeed I had no output, but maybe I used it wrong or maybe the behavior was reverted in Python 3.5 (because it clearly contradicts the documentation you referenced).

Ok so maybe we can add something like that in ._utils.py:

class TqdmDeprecationWarning(object):
    @staticmethod
    def warn(message, file=sys.stderr):
        file.write('DeprecationWarning: %s' % message)

Then to use:

from ._utils import TqdmDeprecationWarning

TqdmDeprecationWarning.warn('parameter xxx should not be used, prefer xxx')

What do you think about it guys?

Coverage decreased (-1.5%) to 79.926% when pulling 7f6c45c on pandas_notebook_fix into 1b4d483 on master.

Coverage increased (+12.5%) to 93.886% when pulling 9fa8cc5 on pandas_notebook_fix into 1b4d483 on master.