Parquet - support for Delta Binary Pcked encoding #1764
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
… to reduce memory consumption
norberttech
commented
Jul 13, 2025
src/lib/parquet/src/Flow/Parquet/Writer/ColumnChunkBuilders.php
Outdated
Show resolved
Hide resolved
Codecov ReportAttention: Patch coverage is
✅ All tests successful. No failed tests found. Additional details and impacted files@@ Coverage Diff @@
## 1.x #1764 +/- ##
==========================================
+ Coverage 81.68% 81.79% +0.10%
==========================================
Files 718 725 +7
Lines 20279 20771 +492
==========================================
+ Hits 16565 16989 +424
- Misses 3714 3782 +68
🚀 New features to boost your workflow:
|
Contributor
Flow PHP - BenchmarksResults of the benchmarks from this PR are compared with the results from 1.x branch. Extractors+-----------------------+------------------------+------+-----+-----------------+------------------+-----------------+
| benchmark | subject | revs | its | mem_peak | mode | rstdev |
+-----------------------+------------------------+------+-----+-----------------+------------------+-----------------+
| CSVExtractorBench | bench_extract_10k | 1 | 3 | 4.865mb +0.08% | 439.630ms +0.85% | ±0.30% -51.34% |
| ExcelExtractorBench | bench_extract_10k_ods | 1 | 3 | 65.560mb +0.01% | 1.065s +1.43% | ±0.45% -56.97% |
| ExcelExtractorBench | bench_extract_10k_xlsx | 1 | 3 | 67.660mb +0.01% | 1.690s -0.42% | ±0.38% -50.93% |
| JsonExtractorBench | bench_extract_10k | 1 | 3 | 5.457mb +0.07% | 1.143s +0.65% | ±0.03% -97.41% |
| ParquetExtractorBench | bench_extract_10k | 1 | 3 | 10.647mb -0.03% | 9.321s -17.99% | ±0.84% +143.97% |
| TextExtractorBench | bench_extract_10k | 1 | 3 | 4.588mb +0.09% | 42.683ms +1.99% | ±1.27% +54.36% |
| XmlExtractorBench | bench_extract_10k | 1 | 3 | 4.573mb +0.09% | 607.992ms +2.22% | ±0.86% -45.52% |
+-----------------------+------------------------+------+-----+-----------------+------------------+-----------------+
Transformers+---------------------------------+--------------------------+------+-----+------------------+-----------------+----------------+
| benchmark | subject | revs | its | mem_peak | mode | rstdev |
+---------------------------------+--------------------------+------+-----+------------------+-----------------+----------------+
| RenameEachEntryTransformerBench | bench_transform_10k_rows | 1 | 3 | 18.584mb +0.02% | 71.826ms -1.50% | ±0.30% -75.91% |
| RenameEntryTransformerBench | bench_transform_10k_rows | 1 | 3 | 123.322mb +0.00% | 66.268ms -2.11% | ±1.14% -33.08% |
+---------------------------------+--------------------------+------+-----+------------------+-----------------+----------------+
Loaders+--------------------+----------------+------+-----+------------------+------------------+----------------+
| benchmark | subject | revs | its | mem_peak | mode | rstdev |
+--------------------+----------------+------+-----+------------------+------------------+----------------+
| CSVLoaderBench | bench_load_10k | 1 | 3 | 62.526mb +0.01% | 83.876ms -3.58% | ±1.59% -24.18% |
| JsonLoaderBench | bench_load_10k | 1 | 3 | 80.607mb +0.00% | 100.408ms -5.25% | ±0.45% -84.68% |
| ParquetLoaderBench | bench_load_10k | 1 | 3 | 835.122mb +0.03% | 18.928s -29.86% | ±0.69% -18.79% |
| TextLoaderBench | bench_load_10k | 1 | 3 | 17.890mb +0.02% | 29.640ms -2.15% | ±0.34% -67.49% |
+--------------------+----------------+------+-----+------------------+------------------+----------------+
Building Blocks+-------------------+----------------------------+------+-----+------------------+------------------+-----------------+
| benchmark | subject | revs | its | mem_peak | mode | rstdev |
+-------------------+----------------------------+------+-----+------------------+------------------+-----------------+
| EntryFactoryBench | bench_entry_factory | 1 | 3 | 106.004mb +0.00% | 644.658ms -1.88% | ±1.20% -36.05% |
| EntryFactoryBench | bench_entry_factory | 1 | 3 | 55.279mb +0.01% | 326.411ms -0.97% | ±2.84% +0.92% |
| EntryFactoryBench | bench_entry_factory | 1 | 3 | 14.865mb +0.03% | 68.935ms -2.60% | ±0.95% -10.12% |
| TypeDetectorBench | bench_type_detector | 1 | 3 | 42.535mb +0.01% | 403.678ms -1.69% | ±1.49% +39.23% |
| TypeDetectorBench | bench_type_detector | 1 | 3 | 11.592mb +0.03% | 81.170ms -1.70% | ±0.35% -80.85% |
| RowsBench | bench_chunk_10_on_10k | 2 | 3 | 93.475mb +0.00% | 3.420ms -7.53% | ±0.34% -75.11% |
| RowsBench | bench_diff_left_1k_on_10k | 2 | 3 | 110.846mb +0.00% | 237.985ms -0.51% | ±0.50% -22.15% |
| RowsBench | bench_diff_right_1k_on_10k | 2 | 3 | 93.566mb +0.00% | 23.679ms -0.30% | ±0.35% +15.16% |
| RowsBench | bench_drop_1k_on_10k | 2 | 3 | 94.350mb +0.00% | 1.506ms -14.66% | ±2.60% -26.56% |
| RowsBench | bench_drop_right_1k_on_10k | 2 | 3 | 94.350mb +0.00% | 1.514ms -9.34% | ±0.56% -78.74% |
| RowsBench | bench_entries_on_10k | 2 | 3 | 92.510mb +0.00% | 3.457ms +1.75% | ±0.50% -84.70% |
| RowsBench | bench_filter_on_10k | 2 | 3 | 93.039mb +0.00% | 16.954ms +9.85% | ±1.85% +50.93% |
| RowsBench | bench_find_on_10k | 2 | 3 | 93.039mb +0.00% | 16.527ms +3.08% | ±1.47% -39.39% |
| RowsBench | bench_find_one_on_10k | 10 | 3 | 91.728mb +0.00% | 1.794μs -5.88% | ±2.67% +9.43% |
| RowsBench | bench_first_on_10k | 10 | 3 | 91.728mb +0.00% | 0.300μs -25.00% | ±0.00% -100.00% |
| RowsBench | bench_flat_map_on_1k | 2 | 3 | 100.789mb +0.00% | 14.569ms -1.25% | ±0.47% -82.66% |
| RowsBench | bench_map_on_10k | 2 | 3 | 130.216mb +0.00% | 68.100ms -2.22% | ±3.21% +214.92% |
| RowsBench | bench_merge_1k_on_10k | 2 | 3 | 93.559mb +0.00% | 1.348ms -2.22% | ±1.37% -51.96% |
| RowsBench | bench_partition_by_on_10k | 2 | 3 | 96.928mb +0.00% | 61.820ms -1.61% | ±0.62% -65.34% |
| RowsBench | bench_remove_on_10k | 2 | 3 | 94.612mb +0.00% | 3.666ms -6.65% | ±1.33% -61.45% |
| RowsBench | bench_sort_asc_on_1k | 2 | 3 | 92.090mb +0.00% | 39.279ms -1.82% | ±1.42% +73.96% |
| RowsBench | bench_sort_by_on_1k | 2 | 3 | 92.090mb +0.00% | 39.311ms -2.32% | ±0.98% +198.05% |
| RowsBench | bench_sort_desc_on_1k | 2 | 3 | 92.090mb +0.00% | 39.701ms -1.31% | ±2.13% +47.75% |
| RowsBench | bench_sort_entries_on_1k | 2 | 3 | 94.171mb +0.00% | 8.047ms -1.86% | ±0.24% -84.05% |
| RowsBench | bench_sort_on_1k | 2 | 3 | 91.921mb +0.00% | 29.222ms -2.05% | ±0.84% -50.79% |
| RowsBench | bench_take_1k_on_10k | 10 | 3 | 91.728mb +0.00% | 13.841μs -13.11% | ±1.82% +23.81% |
| RowsBench | bench_take_right_1k_on_10k | 10 | 3 | 91.728mb +0.00% | 15.543μs -5.99% | ±2.11% -14.58% |
| RowsBench | bench_unique_on_1k | 2 | 3 | 110.847mb +0.00% | 240.563ms -0.59% | ±0.13% -80.10% |
+-------------------+----------------------------+------+-----+------------------+------------------+-----------------+
Parquet Library+--------------------+---------------------------------+------+-----+----------------+-------------------+------------------+
| benchmark | subject | revs | its | mem_peak | mode | rstdev |
+--------------------+---------------------------------+------+-----+----------------+-------------------+------------------+
| ParquetReaderBench | bench_page_headers | 1 | 3 | 6.656mb +0.07% | 3.339s +0.70% | ±0.84% -29.29% |
| ParquetReaderBench | bench_read_metadata | 1 | 3 | 5.341mb +0.08% | 18.055ms -0.07% | ±0.37% -25.06% |
| ParquetReaderBench | bench_read_schema | 1 | 3 | 5.341mb +0.08% | 18.122ms -0.32% | ±0.27% +18.84% |
| ParquetReaderBench | bench_read_values_all_columns | 1 | 3 | 9.079mb -0.03% | 5.657s -28.22% | ±0.68% -21.93% |
| ParquetReaderBench | bench_read_values_single_column | 1 | 3 | 6.377mb -0.05% | 229.631ms -50.28% | ±0.83% +38.75% |
| ParquetReaderBench | bench_read_values_with_limit | 1 | 3 | 6.907mb -0.22% | 28.849ms -14.23% | ±0.28% +27.19% |
| ParquetWriterBench | bench_write_batch | 1 | 3 | 9.823mb -3.46% | 164.869ms -17.17% | ±0.68% +4186.18% |
| ParquetWriterBench | bench_write_gzip | 1 | 3 | 9.785mb +0.10% | 178.742ms +0.88% | ±1.40% +110.14% |
| ParquetWriterBench | bench_write_row_by_row | 1 | 3 | 9.823mb -3.46% | 165.357ms -16.70% | ±1.11% +106.85% |
| ParquetWriterBench | bench_write_snappy | 1 | 3 | 9.823mb -3.46% | 164.906ms -17.58% | ±0.12% -41.53% |
| ParquetWriterBench | bench_write_uncompressed | 1 | 3 | 9.600mb +0.11% | 163.845ms -0.13% | ±1.25% +366.20% |
+--------------------+---------------------------------+------+-----+----------------+-------------------+------------------+
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Resolves: #697
Change Log
Added
Fixed
Changed
Removed
Deprecated
Security
Delta Binary Packed is applicable to Int32 and Int64 (which means that it also applies to Timestamps and other logical types stored as int32/64).
The goal is to significantly reduce output file size through storing only delta between integers in a column rather than integers themself.
This is super useful when we are dealing with incremental datasets, like for example orders.