GH-45227: [C++][Parquet] Enable Size Stats and Page Index by default

[wgtmac]

Rationale for this change

Benchmark data shows that enabling page index and size stats by default does not have significant penalty.

What changes are included in this PR?

Enable the parquet writer to generate page index and size stats by default.

Are these changes tested?

Pass CIs.

Are there any user-facing changes?

No.

• GitHub Issue: [C++][Parquet] Enable size statistics by default #45227

[github-actions]

⚠️ GitHub issue #45227has been automatically assigned in GitHub to PR creator.

[wgtmac]

@pitrou@mapleFU Would you please take a look?

[mapleFU]

Looks ok to me but I don't know what other think about this, do we report need this in maillist?

[pitrou]

Looks ok to me but I don't know what other think about this, do we report need this in maillist?

We can probably mention it on the dev@parquet ML?

[wgtmac]

For the record, the benchmark data on my dev box:

Run on (32 X 2700 MHz CPU s)
CPU Caches:
L1 Data 48 KiB (x16)
L1 Instruction 32 KiB (x16)
L2 Unified 1280 KiB (x16)
L3 Unified 49152 KiB (x1)
Load Average: 1.01, 1.06, 0.60
------------------------------------------------------------------------------------------------------------------------------------------------------
Benchmark Time CPU Iterations UserCounters...
------------------------------------------------------------------------------------------------------------------------------------------------------
BM_WritePrimitiveColumn<SizeStatisticsLevel::None, ::arrow::Int64Type, false> 9386745 ns 9386625 ns 74 bytes_per_second=865.593Mi/s items_per_second=111.71M/s output_size=546.08k page_index_size=0
BM_WritePrimitiveColumn<SizeStatisticsLevel::None, ::arrow::Int64Type, true> 9387974 ns 9387716 ns 75 bytes_per_second=865.493Mi/s items_per_second=111.697M/s output_size=546.091k page_index_size=33
BM_WritePrimitiveColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::Int64Type, true> 9392147 ns 9391889 ns 74 bytes_per_second=865.108Mi/s items_per_second=111.647M/s output_size=546.107k page_index_size=33
BM_WritePrimitiveColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::Int64Type, true> 9363539 ns 9363364 ns 75 bytes_per_second=867.744Mi/s items_per_second=111.987M/s output_size=546.121k page_index_size=47
BM_WritePrimitiveColumn<SizeStatisticsLevel::None, ::arrow::StringType, false> 12773243 ns 12772995 ns 55 bytes_per_second=362.473Mi/s items_per_second=82.0932M/s output_size=864.052k page_index_size=0
BM_WritePrimitiveColumn<SizeStatisticsLevel::None, ::arrow::StringType, true> 12777189 ns 12776940 ns 55 bytes_per_second=362.361Mi/s items_per_second=82.0678M/s output_size=864.083k page_index_size=30
BM_WritePrimitiveColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::StringType, true> 12826700 ns 12826328 ns 54 bytes_per_second=360.965Mi/s items_per_second=81.7518M/s output_size=864.103k page_index_size=30
BM_WritePrimitiveColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::StringType, true> 12837516 ns 12837345 ns 54 bytes_per_second=360.656Mi/s items_per_second=81.6817M/s output_size=864.122k page_index_size=44
BM_WriteListColumn<SizeStatisticsLevel::None, ::arrow::Int64Type, false> 16068608 ns 16068385 ns 43 bytes_per_second=531.323Mi/s items_per_second=65.2571M/s output_size=625.904k page_index_size=0
BM_WriteListColumn<SizeStatisticsLevel::None, ::arrow::Int64Type, true> 16026976 ns 16026766 ns 44 bytes_per_second=532.703Mi/s items_per_second=65.4266M/s output_size=625.915k page_index_size=34
BM_WriteListColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::Int64Type, true> 16723670 ns 16723159 ns 42 bytes_per_second=510.52Mi/s items_per_second=62.702M/s output_size=625.937k page_index_size=34
BM_WriteListColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::Int64Type, true> 16784041 ns 16783691 ns 41 bytes_per_second=508.678Mi/s items_per_second=62.4759M/s output_size=625.957k page_index_size=54
BM_WriteListColumn<SizeStatisticsLevel::None, ::arrow::StringType, false> 19524874 ns 19524500 ns 36 bytes_per_second=258.258Mi/s items_per_second=53.7057M/s output_size=944.092k page_index_size=0
BM_WriteListColumn<SizeStatisticsLevel::None, ::arrow::StringType, true> 19822498 ns 19822084 ns 35 bytes_per_second=254.381Mi/s items_per_second=52.8994M/s output_size=944.123k page_index_size=31
BM_WriteListColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::StringType, true> 20485041 ns 20484509 ns 34 bytes_per_second=246.155Mi/s items_per_second=51.1887M/s output_size=944.149k page_index_size=31
BM_WriteListColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::StringType, true> 20354835 ns 20354291 ns 34 bytes_per_second=247.73Mi/s items_per_second=51.5162M/s output_size=944.174k page_index_size=51

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit 1fcc892.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 14 possible false positives for unstable benchmarks that are known to sometimes produce them.