search

flow-php / flow

Flow PHP - data processing framework
https://flow-php.com
MIT License
491 stars 28 forks source link

Simplified caching pipeline batch size configuration #1239

Closed norberttech closed 1 month ago

norberttech commented 1 month ago

Change Log

Added

Fixed

Changed

  • CachingPipeline will relay only on previously set batch size without modifying it

Removed

  • Default global CachingPipeline batch size

Deprecated

Security

Description

Previously using cache was changing the batch size which for long and complicated pipelines could end up in significant performance drop.

Before

df()
  ->read(...)
  ->batchSize(10_000)
  ->withEntry(...) // batch size here is 10k
  ->cache("test") // cache batch size is not provided so default batch size is used and changed to 100
  ->withEntry(...) // batch size here is 100 
  ->run()

df()
  ->read(...)
  ->batchSize(10_000)
  ->withEntry(...) // batch size here is 10k
  ->cache("test", cacheBatchSize: 10) // cache batch size is set to 10
  ->withEntry(...) // batch size here is 10
  ->run()

After

df()
  ->read(...)
  ->batchSize(10_000)
  ->withEntry(...) // batch size here is 10k
  ->cache("test") // batch size here is still 10k 
  ->withEntry(...) // batch size here is still 10k
  ->run()

df()
  ->read(...)
  ->batchSize(10_000)
  ->withEntry(...) // batch size here is 10k
  ->cache("test", cacheBatchSize: 5000) // batch size here is still 5k
  ->withEntry(...) // batch size here is now 5k
  ->run()
github-actions[bot] commented 1 month ago

Flow PHP - Benchmarks

Results of the benchmarks from this PR are compared with the results from 1.x branch.

Extractors ```shell +-----------------------+-------------------+------+-----+-----------------+------------------+-----------------+ | benchmark | subject | revs | its | mem_peak | mode | rstdev | +-----------------------+-------------------+------+-----+-----------------+------------------+-----------------+ | CSVExtractorBench | bench_extract_10k | 1 | 3 | 4.631mb -0.13% | 523.131ms +4.21% | ±0.86% -39.48% | | JsonExtractorBench | bench_extract_10k | 1 | 3 | 4.719mb -0.13% | 1.119s +3.55% | ±2.32% +348.66% | | ParquetExtractorBench | bench_extract_10k | 1 | 3 | 29.175mb -0.02% | 459.197ms +7.86% | ±0.56% +110.90% | | TextExtractorBench | bench_extract_10k | 1 | 3 | 4.360mb -0.14% | 33.619ms +2.12% | ±1.98% -6.99% | | XmlExtractorBench | bench_extract_10k | 1 | 3 | 4.341mb -0.14% | 657.414ms +3.80% | ±0.23% -26.21% | +-----------------------+-------------------+------+-----+-----------------+------------------+-----------------+ ```
Transformers ```shell +-----------------------------+--------------------------+------+-----+------------------+-----------------+---------------+ | benchmark | subject | revs | its | mem_peak | mode | rstdev | +-----------------------------+--------------------------+------+-----+------------------+-----------------+---------------+ | RenameEntryTransformerBench | bench_transform_10k_rows | 1 | 3 | 116.701mb -0.06% | 62.846ms +4.83% | ±1.29% -1.02% | +-----------------------------+--------------------------+------+-----+------------------+-----------------+---------------+ ```
Loaders ```shell +--------------------+----------------+------+-----+------------------+------------------+----------------+ | benchmark | subject | revs | its | mem_peak | mode | rstdev | +--------------------+----------------+------+-----+------------------+------------------+----------------+ | CSVLoaderBench | bench_load_10k | 1 | 3 | 54.829mb -0.01% | 123.920ms +7.67% | ±0.78% -58.16% | | JsonLoaderBench | bench_load_10k | 1 | 3 | 90.412mb -0.01% | 103.357ms +6.50% | ±0.55% -68.65% | | ParquetLoaderBench | bench_load_10k | 1 | 3 | 124.532mb -0.00% | 1.249s +2.89% | ±0.48% -66.07% | | TextLoaderBench | bench_load_10k | 1 | 3 | 17.551mb -0.04% | 29.605ms +12.00% | ±1.01% -63.99% | +--------------------+----------------+------+-----+------------------+------------------+----------------+ ```
Building Blocks ```shell +-------------------------+----------------------------+------+-----+------------------+------------------+-----------------+ | benchmark | subject | revs | its | mem_peak | mode | rstdev | +-------------------------+----------------------------+------+-----+------------------+------------------+-----------------+ | TypeDetectorBench | bench_type_detector | 1 | 3 | 53.271mb +0.00% | 401.419ms +4.56% | ±0.76% -20.83% | | TypeDetectorBench | bench_type_detector | 1 | 3 | 13.537mb +0.00% | 80.742ms +0.63% | ±0.18% -94.87% | | NativeEntryFactoryBench | bench_entry_factory | 1 | 3 | 107.468mb +0.00% | 490.784ms +6.38% | ±0.42% -72.38% | | NativeEntryFactoryBench | bench_entry_factory | 1 | 3 | 55.826mb +0.00% | 245.336ms +2.29% | ±0.70% -14.17% | | NativeEntryFactoryBench | bench_entry_factory | 1 | 3 | 14.664mb +0.00% | 54.374ms +9.58% | ±1.34% -51.75% | | RowsBench | bench_chunk_10_on_10k | 2 | 3 | 87.445mb +0.00% | 3.543ms +6.06% | ±1.22% -55.84% | | RowsBench | bench_diff_left_1k_on_10k | 2 | 3 | 102.983mb +0.00% | 191.229ms +1.82% | ±0.64% -60.27% | | RowsBench | bench_diff_right_1k_on_10k | 2 | 3 | 85.703mb +0.00% | 19.242ms +1.18% | ±0.71% -66.15% | | RowsBench | bench_drop_1k_on_10k | 2 | 3 | 88.685mb +0.00% | 2.239ms +34.92% | ±1.38% -43.84% | | RowsBench | bench_drop_right_1k_on_10k | 2 | 3 | 88.685mb +0.00% | 2.328ms +38.03% | ±0.88% -42.32% | | RowsBench | bench_entries_on_10k | 2 | 3 | 85.797mb +0.00% | 3.046ms +15.38% | ±2.36% +14.23% | | RowsBench | bench_filter_on_10k | 2 | 3 | 86.326mb +0.00% | 18.749ms +22.13% | ±0.93% +39.69% | | RowsBench | bench_find_on_10k | 2 | 3 | 86.326mb +0.00% | 19.121ms +19.92% | ±0.89% -29.64% | | RowsBench | bench_find_one_on_10k | 10 | 3 | 84.230mb +0.00% | 1.706μs +13.74% | ±2.72% +0.00% | | RowsBench | bench_first_on_10k | 10 | 3 | 84.230mb +0.00% | 0.400μs +33.33% | ±0.00% +0.00% | | RowsBench | bench_flat_map_on_1k | 2 | 3 | 93.580mb +0.00% | 12.683ms +0.14% | ±0.42% -82.81% | | RowsBench | bench_map_on_10k | 2 | 3 | 122.951mb +0.00% | 61.294ms +3.34% | ±1.41% -45.18% | | RowsBench | bench_merge_1k_on_10k | 2 | 3 | 86.846mb +0.00% | 1.698ms +36.17% | ±3.44% +51.75% | | RowsBench | bench_partition_by_on_10k | 2 | 3 | 90.136mb +0.00% | 59.170ms +4.63% | ±0.41% -43.89% | | RowsBench | bench_remove_on_10k | 2 | 3 | 88.947mb +0.00% | 4.213ms +6.25% | ±1.23% -46.20% | | RowsBench | bench_sort_asc_on_1k | 2 | 3 | 84.314mb +0.00% | 41.837ms +8.66% | ±2.75% +1.31% | | RowsBench | bench_sort_by_on_1k | 2 | 3 | 84.314mb +0.00% | 40.774ms +7.07% | ±0.35% -69.23% | | RowsBench | bench_sort_desc_on_1k | 2 | 3 | 84.314mb +0.00% | 41.704ms +7.42% | ±1.26% +0.48% | | RowsBench | bench_sort_entries_on_1k | 2 | 3 | 86.671mb +0.00% | 7.542ms +8.40% | ±0.83% +126.83% | | RowsBench | bench_sort_on_1k | 2 | 3 | 84.230mb +0.00% | 29.240ms +6.67% | ±0.70% -56.67% | | RowsBench | bench_take_1k_on_10k | 10 | 3 | 84.230mb +0.00% | 13.780μs +13.30% | ±1.23% -60.72% | | RowsBench | bench_take_right_1k_on_10k | 10 | 3 | 84.230mb +0.00% | 15.882μs +1.77% | ±0.90% +196.84% | | RowsBench | bench_unique_on_1k | 2 | 3 | 102.988mb +0.00% | 195.424ms +1.72% | ±0.91% -2.37% | +-------------------------+----------------------------+------+-----+------------------+------------------+-----------------+ ```