A native parquet reader for primitive types by copperybean · Pull Request #60361 · ClickHouse/ClickHouse

copperybean · 2024-02-23T16:43:55Z

Changelog category (leave one):

Performance Improvement

Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):

Add a native parquet reader, which can read parquet binary to ClickHouse Columns directly. It's controlled by the setting input_format_parquet_use_native_reader (disabled by default)

Currently, parquet file is read by arrow library, and it's read to arrow table first, and then copy the arrow table to ClickHouse Columns. There are some shortcomings in performance.

1st, it's a redundant operation of copying data from arrow table to ClickHouse Columns.
2nd, in arrow's implementation, repetition and definition levels are read to int16_t array first; then nullability related valid_bits_ is generated. While, the null_map and offsets (not included this time) can be generated directly when reading definition and repetition levels.
3rd, for some situation, the arrow's implementation is inefficient, such as DataTypeDecimal32 type and dictionary parquet row group.

This feature is first implemented in the product BMR of Baidu AI Cloud, which has been fully tested.

Performance Test

As a result, the perforation of current implementation is speedup obviously. To generate the test data, a parquet file src.parquet of TPCDS table store_sales with scale 5000 is used, there are 35207247 rows in this file. Next, the test data is generated with following query:

select
  ss_item_sk,
  ss_wholesale_cost,
  ss_sales_price * ss_item_sk * ss_customer_sk as ss_sales_price_,
  concat('test string with ', toString(ss_item_sk % 1000)) as dict_str,
  concat('test string with ', toString(ss_item_sk * ss_cdemo_sk * ss_hdemo_sk)) as normal_str
from file('<path>/src.parquet', 'Parquet', 'ss_item_sk Nullable(Int32), ss_customer_sk Nullable(Int32), ss_cdemo_sk Nullable(Int32), ss_hdemo_sk Nullable(Int32), ss_wholesale_cost Nullable(Decimal(7, 2)), ss_sales_price Nullable(Decimal(21,6))')
into outfile '<path>/test.parquet' format Parquet

Then the performance is tested by following command

clickhouse-local --query "select max(<field-name>) from file('<path>/test.parquet', 'Parquet')" \
  --max_parsing_threads --max_threads 1 --input_format_parquet_use_native_reader <flag>

For each field, two types tests are triggered with different input_format_parquet_use_native_reader setting, and single thread is used. The parquet reading duration is counted as commit log duration while reading parquet. The CPU model used by this test is Intel(R) Xeon(R) Gold 6271C CPU @ 2.60GHz.

The test result is detailed in following table

Column	Current Duration	Previous Duration	SpeedUp
ss_item_sk Nullable(Int32)	117ms	141ms	1.2
ss_wholesale_cost Nullable(Decimal(7, 2))	341ms	1132ms	3.3
ss_sales_price_ Nullable(Decimal(21, 6))	720ms	1185ms	1.6
dict_str Nullable(String)	259ms	624ms	2.4
normal_str Nullable(String)	717ms	1154ms	1.6

Documentation entry for user-facing changes

Documentation is written (mandatory for new features)

Information about CI checks: https://clickhouse.com/docs/en/development/continuous-integration/

clickhouse-ci · 2024-02-23T16:47:44Z

This is an automatic comment. The PR descriptions does not match the template.

Please, edit it accordingly.

The error is: More than one changelog category specified: 'Improvement', 'Performance Improvement'

robot-clickhouse · 2024-02-23T17:38:17Z

This is an automated comment for commit 3d7befe with description of existing statuses. It's updated for the latest CI running

❌ Click here to open a full report in a separate page

Check name	Description	Status
A Sync	There's no description for the check yet, please add it to tests/ci/ci_config.py:CHECK_DESCRIPTIONS	⏳ pending
CI running	A meta-check that indicates the running CI. Normally, it's in success or pending state. The failed status indicates some problems with the PR	⏳ pending
Integration tests	The integration tests report. In parenthesis the package type is given, and in square brackets are the optional part/total tests	❌ failure
Mergeable Check	Checks if all other necessary checks are successful	❌ failure
Stateless tests	Runs stateless functional tests for ClickHouse binaries built in various configurations -- release, debug, with sanitizers, etc	❌ failure

Successful checks

Check name	Description	Status
AST fuzzer	Runs randomly generated queries to catch program errors. The build type is optionally given in parenthesis. If it fails, ask a maintainer for help	✅ success
ClickBench	Runs [ClickBench](https://github.com/ClickHouse/ClickBench/) with instant-attach table	✅ success
ClickHouse build check	Builds ClickHouse in various configurations for use in further steps. You have to fix the builds that fail. Build logs often has enough information to fix the error, but you might have to reproduce the failure locally. The cmake options can be found in the build log, grepping for cmake. Use these options and follow the general build process	✅ success
Compatibility check	Checks that clickhouse binary runs on distributions with old libc versions. If it fails, ask a maintainer for help	✅ success
Docker keeper image	The check to build and optionally push the mentioned image to docker hub	✅ success
Docker server image	The check to build and optionally push the mentioned image to docker hub	✅ success
Docs check	Builds and tests the documentation	✅ success
Fast test	Normally this is the first check that is ran for a PR. It builds ClickHouse and runs most of stateless functional tests, omitting some. If it fails, further checks are not started until it is fixed. Look at the report to see which tests fail, then reproduce the failure locally as described here	✅ success
Flaky tests	Checks if new added or modified tests are flaky by running them repeatedly, in parallel, with more randomization. Functional tests are run 100 times with address sanitizer, and additional randomization of thread scheduling. Integrational tests are run up to 10 times. If at least once a new test has failed, or was too long, this check will be red. We don't allow flaky tests, read the doc	✅ success
Install packages	Checks that the built packages are installable in a clear environment	✅ success
PR Check	There's no description for the check yet, please add it to tests/ci/ci_config.py:CHECK_DESCRIPTIONS	✅ success
Performance Comparison	Measure changes in query performance. The performance test report is described in detail here. In square brackets are the optional part/total tests	✅ success
Stateful tests	Runs stateful functional tests for ClickHouse binaries built in various configurations -- release, debug, with sanitizers, etc	✅ success
Stress test	Runs stateless functional tests concurrently from several clients to detect concurrency-related errors	✅ success
Style check	Runs a set of checks to keep the code style clean. If some of tests failed, see the related log from the report	✅ success
Unit tests	Runs the unit tests for different release types	✅ success
Upgrade check	Runs stress tests on server version from last release and then tries to upgrade it to the version from the PR. It checks if the new server can successfully startup without any errors, crashes or sanitizer asserts	✅ success

copperybean · 2024-02-29T01:57:03Z

Well, I think these break tests are not caused by my commit. UnitTestsAsan and ASTFuzzerTestAsan have succeeded in former test. While the integration test is failed because of logical error.

Hope for further suggestion.

alexey-milovidov · 2024-02-29T02:43:15Z

Yes, we need to start the investigation, and then check and fix every failure one by one.

copperybean · 2024-03-13T09:58:04Z

Yes, we need to start the investigation, and then check and fix every failure one by one.

@alexey-milovidov
About the investigation, do you refer to the investigation about this PR, or just about the failure tests?
And, should I rebase to master and trigger the checks again?

alexey-milovidov · 2024-04-27T01:01:53Z

@copperybean, please resolve conflicts.

src/Processors/Formats/Impl/Parquet/ParquetRecordReader.cpp

al13n321

Cool, thanks for working on this!

Lots of review comments, mostly unimportant. The only blockers are:

Seems that TIMESTAMP types will be incorrectly interpreted as seconds, please test.
The block_missing_values_ptr thing, please test. If it doesn't work, at least add a TODO so we don't forget.
Pass file metadata to ParquetFileReader::Open to avoid re-reading it for each row group.

tests/queries/0_stateless/02998_native_parquet_reader.sh

al13n321 · 2024-05-02T23:53:20Z

src/Processors/Formats/Impl/ParquetBlockInputFormat.cpp

+        /// If defaults_for_omitted_fields is true, calculate the default values from default expression for omitted fields.
+        /// Otherwise fill the missing columns with zero values of its type.
+        BlockMissingValues * block_missing_values_ptr = format_settings.defaults_for_omitted_fields ? &res.block_missing_values : nullptr;
+        row_group_batch.arrow_column_to_ch_column->arrowTableToCHChunk(res.chunk, *tmp_table, (*tmp_table)->num_rows(), block_missing_values_ptr);


As this comment says, block_missing_values_ptr is used for applying default expressions to omitted fields (here "omitted" means NULLs, I think).

Native reader doesn't populate block_missing_values_ptr, so it'll probably break default expressions. Please test this:

create table a (x Int64, y String default 'asd') engine Memory insert into function file('t.parquet') select 1 as x, null::Nullable(String) as y insert into a select * from file('t.parquet') select * from a ┌─x─┬─y───┐ 1. │ 1 │ asd │ └───┴─────┘

If it doesn't work, either populate block_missing_values_ptr based on null masks or at least add a TODO comment about that.

Sorry, I missed this feature. I would like to add a TODO comment, because the block-missing-values logic will be modified again when supporting nested type, so the it can be implemented later for all types.

src/Processors/Formats/Impl/Parquet/ParquetRecordReader.cpp

src/Processors/Formats/Impl/Parquet/ParquetDataBuffer.h

src/Processors/Formats/Impl/Parquet/ParquetDataValuesReader.cpp

al13n321 · 2024-05-03T04:21:20Z

src/Processors/Formats/Impl/Parquet/ParquetDataValuesReader.h

+        IndividualVisitor && individual_visitor,
+        RepeatedVisitor && repeated_visitor);


A simpler way to handle nulls is to read all non-null values into a column, then use IColumn::expand to insert default values according to null mask. A little slower because of copying.

I don't mind the current approach, just mentioning the alternative in case it turns out better for the general case of arbitrarily nested arrays, nullables, and tuples.

OK, I'd prefer to keep current implementation currently, may be I will change the logic as your suggestion when supporting nested columns.

al13n321 · 2024-05-03T04:42:45Z

src/Processors/Formats/Impl/Parquet/ParquetLeafColReader.cpp

+    visitColStrIndexType(dictionary->size(), [&]<typename TColVec>(TColVec *)
+    {
+        const TColVec & col_src = *static_cast<const TColVec *>(col_existing.get());
+        reserveColumnStrRows(column, reading_rows_num);
+
+        col_dest.getOffsets().resize(col_src.size());
+        for (size_t i = 0; i < col_src.size(); i++)
+        {
+            auto src_idx = col_src.getData()[i];
+            if (0 == src_idx)
+            {
+                null_map->setNull(i);
+            }
+            auto dict_chars_cursor = col_dict_str.getOffsets()[src_idx - 1];
+            auto str_len = col_dict_str.getOffsets()[src_idx] - dict_chars_cursor;
+            auto dst_chars_cursor = col_dest.getChars().size();
+            col_dest.getChars().resize(dst_chars_cursor + str_len);
+
+            memcpySmallAllowReadWriteOverflow15(
+                &col_dest.getChars()[dst_chars_cursor], &col_dict_str.getChars()[dict_chars_cursor], str_len);
+            col_dest.getOffsets()[i] = col_dest.getChars().size();
+        }
+    });


Would it be easier to create a ColumnLowCardinality and call convertToFullColumn() on it?

Well, it will be easier. But the current implementation is more efficient, even not very much. Because the size of new Column generated by convertToFullColumn is smaller than reading_rows_num, so it should be resized to reading_rows_num again.

By the way, I will implement the logic as your suggestion next time. For current codes, I'd prefer not to change it right now. Instead, I can add some comments.

src/Processors/Formats/Impl/Parquet/ParquetDataValuesReader.cpp

copperybean · 2024-05-06T08:15:38Z

Lots of review comments, mostly unimportant. The only blockers are:

Thanks for the comments, I will fix them in this weekend.

Change-Id: I83a8ec8271edefcd96cb5b3bcd12f6b545d9dec0

Change-Id: I57f025b17a04e2c5dded3f18e7f477841287a2c2

Change-Id: I38b8368b022263d9a71cb3f3e9fdad5d6ca26753

Change-Id: If79741b7456667a8dde3e355d9dc684c2dd84f4f

This reverts commit 5df94b7f8955b541ae37e4bbdc13a1fec9ddbbd9.

Change-Id: I53ade40ba24a742a21f9e09dbab7fff90b032b4b

Change-Id: I8f7ebd173558b16d94d3161cb0b5300e7e78833d

Change-Id: Ia7dbf1d762f7f054a9aa677caaaff6bfe1a42c38

al13n321 · 2024-05-20T02:19:29Z

Fix for the error message:

--- a/src/Processors/Formats/Impl/Parquet/ParquetRecordReader.cpp
+++ b/src/Processors/Formats/Impl/Parquet/ParquetRecordReader.cpp
@@ -312,16 +312,28 @@ ParquetRecordReader::ParquetRecordReader(
 {
     log = &Poco::Logger::get("ParquetRecordReader");
 
+    std::unordered_map<String, parquet::schema::NodePtr> parquet_columns;
+    auto root = file_reader->metadata()->schema()->group_node();
+    for (int i = 0; i < root->field_count(); ++i)
+    {
+        auto & node = root->field(i);
+        parquet_columns.emplace(node->name(), node);
+    }
+
     parquet_col_indice.reserve(header.columns());
     column_readers.reserve(header.columns());
     for (const auto & col_with_name : header)
     {
-        auto idx = file_reader->metadata()->schema()->ColumnIndex(col_with_name.name);
-        if (idx < 0)
-        {
-            auto msg = PreformattedMessage::create("can not find column with name: {}", col_with_name.name);
-            throw Exception(std::move(msg), ErrorCodes::PARQUET_EXCEPTION);
-        }
+        auto it = parquet_columns.find(col_with_name.name);
+        if (it == parquet_columns.end())
+            throw Exception(ErrorCodes::PARQUET_EXCEPTION, "no column with '{}' in parquet file", col_with_name.name);
+
+        auto node = it->second;
+        if (!node->is_primitive())
+            throw Exception(ErrorCodes::NOT_IMPLEMENTED, "arrays and maps are not implemented in native parquet reader");
+
+        auto idx = file_reader->metadata()->schema()->ColumnIndex(*node);
+        chassert(idx >= 0);
         parquet_col_indice.push_back(idx);
     }
     if (reader_properties.pre_buffer())

Clark0 · 2024-05-20T04:03:42Z

src/Processors/Formats/Impl/Parquet/ParquetRecordReader.cpp

+        case parquet::Type::BYTE_ARRAY:
+            return fromByteArray();
+        case parquet::Type::FIXED_LEN_BYTE_ARRAY:
+            return fromFLBA();


It appears that the Bool and FLBA (fixed string) types are not currently supported, Are there any plans to add support for these types in the future?

Bool types are distinct due to their 1-bit width, while clickhouse uses a UInt8 column vector

With current framework, supporting these types is not difficult. My plan is to support nested type after this PR, then support all parquet types and enable 02735_parquet_encoder test.

@copperybean Any news on boolean support? I was looking at the code and it looks like ParquetPlainValuesReader needs to be extended for boolean type. The problem is ParquetPlainValuesReader is tightly coupled to ParquetDataBuffer, which does not allow single bit reads.

Moreover, ParquetPlainValuesReader seems to be templated on the output clickhouse column type. Isn't that a problem? What if there are two distinct parquet types that would result in the same output column type, how would one differentiate? For instance: uint8_t and booleans.

I might be missing something, though. I only had a very superficial read of the code.

Change-Id: Iefec91bca223eedaabe302b7891808c6d94eed9d

Change-Id: Iff9f5f894e815b184ac35f61b4cac87908c612b5

copperybean · 2024-05-22T05:02:10Z

It seems that the ClickHouse special build check is failed because of OOM?

al13n321 · 2024-05-22T05:51:11Z

It's broken in master, waiting for #64204 (and maybe something else, that one only fixes one of the two OOMing builds).

Algunenano · 2024-05-24T11:05:53Z

src/Core/SettingsChangesHistory.h

              {"cross_join_min_bytes_to_compress", 0, 1_GiB, "A new setting."},
              {"http_max_chunk_size", 0, 0, "Internal limitation"},
              {"prefer_external_sort_block_bytes", 0, DEFAULT_BLOCK_SIZE * 256, "Prefer maximum block bytes for external sort, reduce the memory usage during merging."},
+              {"input_format_parquet_use_native_reader", false, false, "When reading Parquet files, to use native reader instead of arrow reader."},


This should go to 24.6

zhanglistar · 2024-06-04T10:09:42Z

What a great feature! @copperybean Will you support complex types? @al13n321 Do you have plan to support complex types to make native parquet reader production ready?

baibaichen · 2024-06-04T10:16:21Z

@copperybean it's cool feature

Do you have plans to support page index? In this case，we need to skip reading some rows.

alexey-milovidov · 2024-07-02T09:03:25Z

I tried a few parquet files from example datasets and found that none worked.

Code: 723. DB::Exception: Unsupported logical type: Int(bitWidth=16, isSigned=false) and physical type: INT32 for field ==date==, bit width: 16: (in file/uri /home/milovidov/uk.parquet): While executing ParquetBlockInputFormat: While executing File. (PARQUET_EXCEPTION)

Code: 48. DB::Exception: arrays and maps are not implemented in native parquet reader: (in file/uri clickhouse-public-datasets/hits.parquet): While executing ParquetBlockInputFormat: While executing S3Source. (NOT_IMPLEMENTED)

alexey-milovidov · 2024-07-02T09:03:56Z

I want to highlight this feature on the release webinar, but so far it looks not ready for use.

alexey-milovidov · 2024-07-02T09:04:41Z

Don't be obsessed with TPC-DS or TPC-H - they are irrelevant in practice. Only academics use these tests.

nikitamikhaylov added the can be tested Allows running workflows for external contributors label Feb 23, 2024

al13n321 self-assigned this Feb 23, 2024

robot-clickhouse added the pr-performance Pull request with some performance improvements label Feb 23, 2024

alexey-milovidov mentioned this pull request Feb 23, 2024

Add Baidu to adopters ClickHouse/clickhouse-docs#2072

Closed

copperybean force-pushed the gcmaster-parquet branch 8 times, most recently from 5012933 to 5166d8b Compare February 28, 2024 06:23

Clark0 reviewed Apr 28, 2024

View reviewed changes

src/Processors/Formats/Impl/Parquet/ParquetRecordReader.cpp Outdated Show resolved Hide resolved

al13n321 requested changes May 3, 2024

View reviewed changes

copperybean added 8 commits May 11, 2024 15:43

original parquet reader

9d0ad7b

Change-Id: I83a8ec8271edefcd96cb5b3bcd12f6b545d9dec0

fix build

8fb89ce

Change-Id: I57f025b17a04e2c5dded3f18e7f477841287a2c2

support reading simple types by native parquet reader

dbdff6c

Change-Id: I38b8368b022263d9a71cb3f3e9fdad5d6ca26753

log duration while reading parquet

8172f6c

Change-Id: If79741b7456667a8dde3e355d9dc684c2dd84f4f

Revert "log duration while reading parquet"

e017915

This reverts commit 5df94b7f8955b541ae37e4bbdc13a1fec9ddbbd9.

add test

18b3ebc

Change-Id: I53ade40ba24a742a21f9e09dbab7fff90b032b4b

fix style

e1fcdba

Change-Id: I8f7ebd173558b16d94d3161cb0b5300e7e78833d

fix test

471dff6

Change-Id: Ia7dbf1d762f7f054a9aa677caaaff6bfe1a42c38

copperybean force-pushed the gcmaster-parquet branch from 5166d8b to 718fa1c Compare May 15, 2024 02:16

Clark0 reviewed May 20, 2024

View reviewed changes

fix reader type, update comment

ad5f6f2

Change-Id: Iefec91bca223eedaabe302b7891808c6d94eed9d

copperybean force-pushed the gcmaster-parquet branch from ddaf172 to ad5f6f2 Compare May 20, 2024 08:47

robot-ch-test-poll2 added the submodule changed At least one submodule changed in this PR. label May 20, 2024

copperybean mentioned this pull request May 21, 2024

Check validation of of bit offset when reading bit packed values ClickHouse/arrow#66

Merged

fix clang-tidy warnings

b253ca3

Change-Id: Iff9f5f894e815b184ac35f61b4cac87908c612b5

copperybean force-pushed the gcmaster-parquet branch from dd95beb to b253ca3 Compare May 21, 2024 08:49

Merge branch 'ClickHouse:master' into gcmaster-parquet

4b1c9ad

al13n321 approved these changes May 22, 2024

View reviewed changes

Merge branch 'master' into gcmaster-parquet

3d7befe

al13n321 added this pull request to the merge queue May 24, 2024

github-merge-queue bot removed this pull request from the merge queue due to no response for status checks May 24, 2024

al13n321 added this pull request to the merge queue May 24, 2024

Merged via the queue into ClickHouse:master with commit ee3e7f2 May 24, 2024

robot-clickhouse added the pr-synced-to-cloud The PR is synced to the cloud repo label May 24, 2024

Algunenano reviewed May 24, 2024

View reviewed changes

Algunenano mentioned this pull request May 24, 2024

Change input_format_parquet_use_native_reader to 24.6 #64359

Merged

nikitamikhaylov mentioned this pull request May 31, 2024

Cleanup the changelog for 24.5 #64704

Merged

22 tasks

chhetripradeep mentioned this pull request Jun 4, 2024

Select query fails with input_format_parquet_use_native_reader=true #64779

Open

lgbo-ustc mentioned this pull request Jun 19, 2024

Use page index to filt out parquet pages in native parquet reader #65428

Closed

19 tasks

		IndividualVisitor && individual_visitor,
		RepeatedVisitor && repeated_visitor);

Conversation

copperybean commented Feb 23, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changelog category (leave one):

Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):

Performance Test

Documentation entry for user-facing changes

Uh oh!

clickhouse-ci bot commented Feb 23, 2024

Uh oh!

robot-clickhouse commented Feb 23, 2024 • edited by robot-ch-test-poll3 Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

copperybean commented Feb 29, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

alexey-milovidov commented Feb 29, 2024

Uh oh!

copperybean commented Mar 13, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

alexey-milovidov commented Apr 27, 2024

Uh oh!

Uh oh!

al13n321 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

copperybean commented May 6, 2024

Uh oh!

al13n321 commented May 20, 2024

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

copperybean commented May 22, 2024

Uh oh!

al13n321 commented May 22, 2024

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

zhanglistar commented Jun 4, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

baibaichen commented Jun 4, 2024

Uh oh!

alexey-milovidov commented Jul 2, 2024

Uh oh!

alexey-milovidov commented Jul 2, 2024

Uh oh!

alexey-milovidov commented Jul 2, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

copperybean commented Feb 23, 2024 •

edited

Loading

robot-clickhouse commented Feb 23, 2024 •

edited by robot-ch-test-poll3

Loading

copperybean commented Feb 29, 2024 •

edited

Loading

copperybean commented Mar 13, 2024 •

edited

Loading

zhanglistar commented Jun 4, 2024 •

edited

Loading

alexey-milovidov commented Jul 2, 2024 •

edited

Loading