Re: [PR] [Parquet] perf: Create StructArrays directly rather than via `ArrayData` [arrow-rs]
alamb commented on code in PR #9120:
URL: https://github.com/apache/arrow-rs/pull/9120#discussion_r2676536711
##
parquet/src/arrow/array_reader/struct_array.rs:
##
@@ -124,16 +124,15 @@ impl ArrayReader for StructArrayReader {
return Err(general_err!("Not all children array length are the
same!"));
}
-// Now we can build array data
-let mut array_data_builder =
ArrayDataBuilder::new(self.data_type.clone())
-.len(children_array_len)
-.child_data(
-children_array
-.into_iter()
-.map(|x| x.into_data())
Review Comment:
converting the child arrays into ArrayData is wasteful for at least 2
reasons:
1. They are just converted back to ArrayRefs below
2. Each array data has at least one new allocation (the Vec of buffers)
##
parquet/src/arrow/array_reader/struct_array.rs:
##
@@ -169,11 +166,18 @@ impl ArrayReader for StructArrayReader {
return Err(general_err!("Failed to decode level data for
struct array"));
}
-array_data_builder =
array_data_builder.null_bit_buffer(Some(bitmap_builder.into()));
+nulls = Some(NullBuffer::new(bitmap_builder.finish()));
Review Comment:
NullBuffer::new counts the set bits, but so do the existing code paths
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
Re: [PR] [Parquet] perf: Create StructArrays directly rather than via `ArrayData` [arrow-rs]
alamb commented on PR #9120: URL: https://github.com/apache/arrow-rs/pull/9120#issuecomment-3730702335 The reason I think this will actually help ClickBench is that the ClickBench dataset has 105 columns. I could probably make a benchmark that shows this helping for reading really wide tables > Read batch with 8192 rows and 105 columns -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
Re: [PR] [Parquet] perf: Create StructArrays directly rather than via `ArrayData` [arrow-rs]
alamb commented on PR #9120: URL: https://github.com/apache/arrow-rs/pull/9120#issuecomment-3730184899 my analysis of results is not a huge win but some improvement -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
Re: [PR] [Parquet] perf: Create StructArrays directly rather than via `ArrayData` [arrow-rs]
alamb-ghbot commented on PR #9120: URL: https://github.com/apache/arrow-rs/pull/9120#issuecomment-3730131472 🤖: Benchmark completed Details ``` groupalamb_less_parquet_allocations main --- arrow_reader_clickbench/async/Q1 1.00 2.3±0.05ms? ?/sec 1.01 2.3±0.01ms? ?/sec arrow_reader_clickbench/async/Q101.00 12.7±0.35ms? ?/sec 1.02 12.9±0.44ms? ?/sec arrow_reader_clickbench/async/Q111.00 14.3±0.33ms? ?/sec 1.04 14.8±0.66ms? ?/sec arrow_reader_clickbench/async/Q121.00 25.5±0.53ms? ?/sec 1.02 26.1±0.55ms? ?/sec arrow_reader_clickbench/async/Q131.00 31.0±0.60ms? ?/sec 1.00 31.0±0.70ms? ?/sec arrow_reader_clickbench/async/Q141.00 28.2±0.68ms? ?/sec 1.01 28.3±0.68ms? ?/sec arrow_reader_clickbench/async/Q191.00 5.2±0.13ms? ?/sec 1.03 5.3±0.14ms? ?/sec arrow_reader_clickbench/async/Q201.22149.2±1.37ms? ?/sec 1.00122.6±1.57ms? ?/sec arrow_reader_clickbench/async/Q211.08169.9±2.38ms? ?/sec 1.00156.6±2.79ms? ?/sec arrow_reader_clickbench/async/Q221.00241.6±2.10ms? ?/sec 1.29 311.7±10.07ms? ?/sec arrow_reader_clickbench/async/Q231.00408.7±2.83ms? ?/sec 1.00406.7±4.25ms? ?/sec arrow_reader_clickbench/async/Q241.00 34.5±0.51ms? ?/sec 1.00 34.7±0.86ms? ?/sec arrow_reader_clickbench/async/Q271.00100.4±1.00ms? ?/sec 1.00100.4±1.43ms? ?/sec arrow_reader_clickbench/async/Q281.01 99.0±0.91ms? ?/sec 1.00 98.4±0.66ms? ?/sec arrow_reader_clickbench/async/Q301.00 30.7±0.58ms? ?/sec 1.01 30.9±0.59ms? ?/sec arrow_reader_clickbench/async/Q361.00108.8±0.91ms? ?/sec 1.00108.9±1.48ms? ?/sec arrow_reader_clickbench/async/Q371.00 84.9±0.61ms? ?/sec 1.01 85.6±1.61ms? ?/sec arrow_reader_clickbench/async/Q381.00 32.7±0.50ms? ?/sec 1.01 33.0±0.48ms? ?/sec arrow_reader_clickbench/async/Q391.00 46.4±0.60ms? ?/sec 1.00 46.3±0.71ms? ?/sec arrow_reader_clickbench/async/Q401.00 27.5±0.60ms? ?/sec 1.00 27.5±0.65ms? ?/sec arrow_reader_clickbench/async/Q411.00 21.8±0.51ms? ?/sec 1.05 23.0±0.64ms? ?/sec arrow_reader_clickbench/async/Q421.00 10.8±0.14ms? ?/sec 1.04 11.2±0.19ms? ?/sec arrow_reader_clickbench/sync/Q1 1.00 2.1±0.01ms? ?/sec 1.03 2.1±0.01ms? ?/sec arrow_reader_clickbench/sync/Q10 1.00 9.9±0.16ms? ?/sec 1.00 9.9±0.11ms? ?/sec arrow_reader_clickbench/sync/Q11 1.00 11.4±0.17ms? ?/sec 1.00 11.4±0.25ms? ?/sec arrow_reader_clickbench/sync/Q12 1.06 36.6±1.80ms? ?/sec 1.00 34.4±1.12ms? ?/sec arrow_reader_clickbench/sync/Q13 1.00 47.9±0.97ms? ?/sec 1.00 47.8±1.07ms? ?/sec arrow_reader_clickbench/sync/Q14 1.00 36.4±0.71ms? ?/sec 1.23 44.6±1.02ms? ?/sec arrow_reader_clickbench/sync/Q19 1.00 4.2±0.03ms? ?/sec 1.03 4.3±0.04ms? ?/sec arrow_reader_clickbench/sync/Q20 1.00177.9±1.11ms? ?/sec 1.00177.8±1.22ms? ?/sec arrow_reader_clickbench/sync/Q21 1.01237.8±2.19ms? ?/sec 1.00235.0±3.23ms? ?/sec arrow_reader_clickbench/sync/Q22 1.00480.8±4.14ms? ?/sec 1.00481.5±4.21ms? ?/sec arrow_reader_clickbench/sync/Q23 1.02 442.3±17.97ms? ?/sec 1.00 433.4±12.76ms? ?/sec arrow_reader_clickbench/sync/Q24 1.00 45.6±1.34ms? ?/sec 1.00 45.8±0.80ms? ?/sec arrow_reader_clickbench/sync/Q27 1.00154.3±1.96ms? ?/sec 1.01155.2±2.01ms? ?/sec arrow_reader_clickbench/sync/Q28 1.00148.8±1.83ms? ?/sec 1.01149.8±2.15ms? ?/sec arrow_reader_clickbench/sync/Q30 1.00 30.9±0.55ms? ?/sec 1.02 31.5±0.97ms? ?/sec arrow_reader_clickbench/sync/Q36 1.00153.3±1.69ms? ?/sec 1.01154.1±1.67ms? ?/sec arrow_reader_clickbench/sync/Q37 1.00 88.4±0.86ms? ?/sec 1.01 89.3±1.68ms? ?/sec arrow_reader_clickbench/sync/Q38 1.00 29.1±0.49ms? ?/sec 1.02 29.6±0.39ms? ?/sec arrow_reader_c
Re: [PR] [Parquet] perf: Create StructArrays directly rather than via `ArrayData` [arrow-rs]
alamb-ghbot commented on PR #9120: URL: https://github.com/apache/arrow-rs/pull/9120#issuecomment-3730046979 🤖 `./gh_compare_arrow.sh` [gh_compare_arrow.sh](https://github.com/alamb/datafusion-benchmarking/blob/main/scripts/gh_compare_arrow.sh) Running Linux aal-dev 6.14.0-1018-gcp #19~24.04.1-Ubuntu SMP Wed Sep 24 23:23:09 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux Comparing alamb/less_parquet_allocations (75ea40bf536af42a1297fc11e5a47b18472b649d) to 96637fc8b928a94de53bbec3501337c0ecfbf936 [diff](https://github.com/apache/arrow-rs/compare/96637fc8b928a94de53bbec3501337c0ecfbf936..75ea40bf536af42a1297fc11e5a47b18472b649d) BENCH_NAME=arrow_reader_clickbench BENCH_COMMAND=cargo bench --features=arrow,async,test_common,experimental --bench arrow_reader_clickbench BENCH_FILTER= BENCH_BRANCH_NAME=alamb_less_parquet_allocations Results will be posted here when complete -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
Re: [PR] [Parquet] perf: Create StructArrays directly rather than via `ArrayData` [arrow-rs]
alamb-ghbot commented on PR #9120: URL: https://github.com/apache/arrow-rs/pull/9120#issuecomment-3730046833 🤖: Benchmark completed Details ``` group alamb_less_parquet_allocations main - -- arrow_array_reader/BYTE_ARRAY/Decimal128Array/plain encoded, mandatory, no NULLs 1.02 1288.5±72.50µs? ?/sec1.00 1269.0±4.39µs? ?/sec arrow_array_reader/BYTE_ARRAY/Decimal128Array/plain encoded, optional, half NULLs 1.00 1297.6±14.79µs? ?/sec1.00 1292.0±16.26µs? ?/sec arrow_array_reader/BYTE_ARRAY/Decimal128Array/plain encoded, optional, no NULLs1.01 1285.3±7.71µs? ?/sec1.00 1276.8±4.46µs? ?/sec arrow_array_reader/BinaryArray/dictionary encoded, mandatory, no NULLs 1.09 519.1±13.71µs? ?/sec1.00 477.7±5.81µs? ?/sec arrow_array_reader/BinaryArray/dictionary encoded, optional, half NULLs 1.01673.1±7.15µs? ?/sec1.00 668.2±29.95µs? ?/sec arrow_array_reader/BinaryArray/dictionary encoded, optional, no NULLs 1.04 509.4±27.48µs? ?/sec1.00 488.1±5.00µs? ?/sec arrow_array_reader/BinaryArray/plain encoded, mandatory, no NULLs 1.04560.7±7.13µs? ?/sec1.00 538.2±9.08µs? ?/sec arrow_array_reader/BinaryArray/plain encoded, optional, half NULLs 1.00727.7±7.03µs? ?/sec1.02 742.8±4.69µs? ?/sec arrow_array_reader/BinaryArray/plain encoded, optional, no NULLs 1.05573.8±8.33µs? ?/sec1.00 547.0±3.26µs? ?/sec arrow_array_reader/BinaryViewArray/dictionary encoded, mandatory, no NULLs 1.00237.6±3.93µs? ?/sec1.05 249.3±2.97µs? ?/sec arrow_array_reader/BinaryViewArray/dictionary encoded, optional, half NULLs 1.05241.5±2.59µs? ?/sec1.00 230.3±3.86µs? ?/sec arrow_array_reader/BinaryViewArray/dictionary encoded, optional, no NULLs 1.00236.5±3.31µs? ?/sec1.08 256.3±2.74µs? ?/sec arrow_array_reader/BinaryViewArray/plain encoded, mandatory, no NULLs 1.01295.2±9.58µs? ?/sec1.00 292.6±2.22µs? ?/sec arrow_array_reader/BinaryViewArray/plain encoded, mandatory, no NULLs, short string1.00 287.8±14.88µs? ?/sec1.06 306.5±4.81µs? ?/sec arrow_array_reader/BinaryViewArray/plain encoded, optional, half NULLs 1.06285.0±3.48µs? ?/sec1.00 268.9±3.39µs? ?/sec arrow_array_reader/BinaryViewArray/plain encoded, optional, no NULLs 1.00302.1±4.74µs? ?/sec1.00 302.5±11.43µs? ?/sec arrow_array_reader/FIXED_LEN_BYTE_ARRAY/Decimal128Array/byte_stream_split encoded, mandatory, no NULLs 1.05 1127.7±6.40µs? ?/sec1.00 1071.6±9.95µs? ?/sec arrow_array_reader/FIXED_LEN_BYTE_ARRAY/Decimal128Array/byte_stream_split encoded, optional, half NULLs1.04 974.9±16.30µs? ?/sec1.00 941.3±20.37µs? ?/sec arrow_array_reader/FIXED_LEN_BYTE_ARRAY/Decimal128Array/byte_stream_split encoded, optional, no NULLs 1.04 1130.6±10.45µs? ?/sec1.00 1089.5±41.68µs? ?/sec arrow_array_reader/FIXED_LEN_BYTE_ARRAY/Decimal128Array/plain encoded, mandatory, no NULLs 1.12454.1±5.67µs? ?/sec1.00 404.1±4.24µs? ?/sec arrow_array_reader/FIXED_LEN_BYTE_ARRAY/Decimal128Array/plain encoded, optional, half NULLs1.09 637.7±10.53µs? ?/sec1.00 586.2±3.94µs? ?/sec arrow_array_reader/FIXED_LEN_BYTE_ARRAY/Decimal128Array/plain encoded, optional, no NULLs 1.11 459.0±14.97µs? ?/sec1.00 415.3±3.92µs? ?/sec arrow_array_reader/FIXED_LEN_BYTE_ARRAY/Float16Array/byte_stream_split encoded, mandatory, no NULLs1.00160.4±2.14µs? ?/sec1.26 202.3±1.43µs? ?/sec arrow_array_reader/FIXED_LEN_BYTE_ARRAY/Float16Array/byte_stream_split encoded, optional, half NULLs 1.00287.4±2.25µs? ?/sec1.10 317.5±1.58µs? ?/sec arrow_array_reader/FIXED_L
Re: [PR] [Parquet] perf: Create StructArrays directly rather than via `ArrayData` [arrow-rs]
alamb-ghbot commented on PR #9120: URL: https://github.com/apache/arrow-rs/pull/9120#issuecomment-3729810850 🤖 `./gh_compare_arrow.sh` [gh_compare_arrow.sh](https://github.com/alamb/datafusion-benchmarking/blob/main/scripts/gh_compare_arrow.sh) Running Linux aal-dev 6.14.0-1018-gcp #19~24.04.1-Ubuntu SMP Wed Sep 24 23:23:09 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux Comparing alamb/less_parquet_allocations (75ea40bf536af42a1297fc11e5a47b18472b649d) to 96637fc8b928a94de53bbec3501337c0ecfbf936 [diff](https://github.com/apache/arrow-rs/compare/96637fc8b928a94de53bbec3501337c0ecfbf936..75ea40bf536af42a1297fc11e5a47b18472b649d) BENCH_NAME=arrow_reader BENCH_COMMAND=cargo bench --features=arrow,async,test_common,experimental --bench arrow_reader BENCH_FILTER= BENCH_BRANCH_NAME=alamb_less_parquet_allocations Results will be posted here when complete -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
Re: [PR] [Parquet] perf: Create StructArrays directly rather than via `ArrayData` [arrow-rs]
alamb-ghbot commented on PR #9120: URL: https://github.com/apache/arrow-rs/pull/9120#issuecomment-3729810577 🤖: Benchmark completed Details ``` groupalamb_less_parquet_allocations main --- arrow_reader_clickbench/async/Q1 1.00 2.3±0.07ms? ?/sec 1.02 2.3±0.09ms? ?/sec arrow_reader_clickbench/async/Q101.00 12.8±0.42ms? ?/sec 1.05 13.3±0.76ms? ?/sec arrow_reader_clickbench/async/Q111.00 14.6±0.55ms? ?/sec 1.02 14.9±0.57ms? ?/sec arrow_reader_clickbench/async/Q121.00 25.9±0.58ms? ?/sec 1.02 26.4±0.70ms? ?/sec arrow_reader_clickbench/async/Q131.02 31.5±0.79ms? ?/sec 1.00 30.9±0.52ms? ?/sec arrow_reader_clickbench/async/Q141.01 28.8±0.65ms? ?/sec 1.00 28.6±0.59ms? ?/sec arrow_reader_clickbench/async/Q191.00 5.2±0.08ms? ?/sec 1.03 5.4±0.22ms? ?/sec arrow_reader_clickbench/async/Q201.21149.2±1.53ms? ?/sec 1.00123.3±0.69ms? ?/sec arrow_reader_clickbench/async/Q211.08170.0±1.53ms? ?/sec 1.00157.7±2.81ms? ?/sec arrow_reader_clickbench/async/Q221.00 308.4±36.14ms? ?/sec 1.01 312.9±12.38ms? ?/sec arrow_reader_clickbench/async/Q231.01410.2±4.15ms? ?/sec 1.00408.1±3.52ms? ?/sec arrow_reader_clickbench/async/Q241.00 34.6±0.94ms? ?/sec 1.02 35.1±0.71ms? ?/sec arrow_reader_clickbench/async/Q271.01101.2±0.65ms? ?/sec 1.00100.5±0.75ms? ?/sec arrow_reader_clickbench/async/Q281.01 99.2±1.48ms? ?/sec 1.00 98.5±0.63ms? ?/sec arrow_reader_clickbench/async/Q301.00 31.1±0.99ms? ?/sec 1.00 31.0±0.63ms? ?/sec arrow_reader_clickbench/async/Q361.00109.9±0.99ms? ?/sec 1.00110.4±4.08ms? ?/sec arrow_reader_clickbench/async/Q371.00 86.0±0.74ms? ?/sec 1.00 85.8±0.76ms? ?/sec arrow_reader_clickbench/async/Q381.00 33.3±0.64ms? ?/sec 1.00 33.2±0.53ms? ?/sec arrow_reader_clickbench/async/Q391.00 46.1±0.46ms? ?/sec 1.01 46.4±0.59ms? ?/sec arrow_reader_clickbench/async/Q401.00 27.3±0.69ms? ?/sec 1.02 27.8±0.73ms? ?/sec arrow_reader_clickbench/async/Q411.00 22.5±0.52ms? ?/sec 1.00 22.6±0.41ms? ?/sec arrow_reader_clickbench/async/Q421.00 11.0±0.20ms? ?/sec 1.01 11.1±0.16ms? ?/sec arrow_reader_clickbench/sync/Q1 1.00 2.1±0.01ms? ?/sec 1.04 2.1±0.06ms? ?/sec arrow_reader_clickbench/sync/Q10 1.00 10.0±0.18ms? ?/sec 1.00 10.0±0.23ms? ?/sec arrow_reader_clickbench/sync/Q11 1.00 11.4±0.25ms? ?/sec 1.01 11.6±0.23ms? ?/sec arrow_reader_clickbench/sync/Q12 1.07 36.9±2.00ms? ?/sec 1.00 34.4±0.61ms? ?/sec arrow_reader_clickbench/sync/Q13 1.00 48.0±1.21ms? ?/sec 1.02 48.7±1.27ms? ?/sec arrow_reader_clickbench/sync/Q14 1.00 44.2±1.11ms? ?/sec 1.02 45.0±1.08ms? ?/sec arrow_reader_clickbench/sync/Q19 1.00 4.2±0.03ms? ?/sec 1.02 4.3±0.03ms? ?/sec arrow_reader_clickbench/sync/Q20 1.00178.7±1.50ms? ?/sec 1.00178.4±1.57ms? ?/sec arrow_reader_clickbench/sync/Q21 1.01237.9±2.33ms? ?/sec 1.00236.5±2.75ms? ?/sec arrow_reader_clickbench/sync/Q22 1.00478.2±4.24ms? ?/sec 1.01484.9±4.94ms? ?/sec arrow_reader_clickbench/sync/Q23 1.02 444.7±18.71ms? ?/sec 1.00 437.1±13.75ms? ?/sec arrow_reader_clickbench/sync/Q24 1.00 46.4±0.90ms? ?/sec 1.01 46.9±0.79ms? ?/sec arrow_reader_clickbench/sync/Q27 1.00154.8±1.48ms? ?/sec 1.00155.4±1.50ms? ?/sec arrow_reader_clickbench/sync/Q28 1.00149.8±1.36ms? ?/sec 1.01151.7±2.33ms? ?/sec arrow_reader_clickbench/sync/Q30 1.00 31.4±0.73ms? ?/sec 1.01 31.6±1.18ms? ?/sec arrow_reader_clickbench/sync/Q36 1.00154.4±2.27ms? ?/sec 1.02156.8±1.59ms? ?/sec arrow_reader_clickbench/sync/Q37 1.00 88.7±0.96ms? ?/sec 1.02 90.6±1.31ms? ?/sec arrow_reader_clickbench/sync/Q38 1.00 29.4±0.64ms? ?/sec 1.01 29.7±0.81ms? ?/sec arrow_reader_c
Re: [PR] [Parquet] perf: Create StructArrays directly rather than via `ArrayData` [arrow-rs]
alamb-ghbot commented on PR #9120: URL: https://github.com/apache/arrow-rs/pull/9120#issuecomment-3729719775 🤖 `./gh_compare_arrow.sh` [gh_compare_arrow.sh](https://github.com/alamb/datafusion-benchmarking/blob/main/scripts/gh_compare_arrow.sh) Running Linux aal-dev 6.14.0-1018-gcp #19~24.04.1-Ubuntu SMP Wed Sep 24 23:23:09 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux Comparing alamb/less_parquet_allocations (75ea40bf536af42a1297fc11e5a47b18472b649d) to 96637fc8b928a94de53bbec3501337c0ecfbf936 [diff](https://github.com/apache/arrow-rs/compare/96637fc8b928a94de53bbec3501337c0ecfbf936..75ea40bf536af42a1297fc11e5a47b18472b649d) BENCH_NAME=arrow_reader_clickbench BENCH_COMMAND=cargo bench --features=arrow,async,test_common,experimental --bench arrow_reader_clickbench BENCH_FILTER= BENCH_BRANCH_NAME=alamb_less_parquet_allocations Results will be posted here when complete -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
Re: [PR] [Parquet] perf: Create StructArrays directly rather than via `ArrayData` [arrow-rs]
alamb-ghbot commented on PR #9120: URL: https://github.com/apache/arrow-rs/pull/9120#issuecomment-3729719550 🤖: Benchmark completed Details ``` group alamb_less_parquet_allocations main - -- arrow_array_reader/BYTE_ARRAY/Decimal128Array/plain encoded, mandatory, no NULLs 1.00 1275.4±14.64µs? ?/sec1.00 1269.4±5.96µs? ?/sec arrow_array_reader/BYTE_ARRAY/Decimal128Array/plain encoded, optional, half NULLs 1.00 1297.5±16.24µs? ?/sec1.00 1291.0±4.78µs? ?/sec arrow_array_reader/BYTE_ARRAY/Decimal128Array/plain encoded, optional, no NULLs1.00 1285.7±14.46µs? ?/sec1.00 1280.7±19.13µs? ?/sec arrow_array_reader/BinaryArray/dictionary encoded, mandatory, no NULLs 1.08 517.9±17.01µs? ?/sec1.00 477.4±11.91µs? ?/sec arrow_array_reader/BinaryArray/dictionary encoded, optional, half NULLs 1.01672.5±5.42µs? ?/sec1.00 665.5±3.98µs? ?/sec arrow_array_reader/BinaryArray/dictionary encoded, optional, no NULLs 1.03504.6±4.95µs? ?/sec1.00 491.6±9.60µs? ?/sec arrow_array_reader/BinaryArray/plain encoded, mandatory, no NULLs 1.04559.1±2.87µs? ?/sec1.00 538.0±10.80µs? ?/sec arrow_array_reader/BinaryArray/plain encoded, optional, half NULLs 1.00728.3±4.37µs? ?/sec1.02 744.3±11.11µs? ?/sec arrow_array_reader/BinaryArray/plain encoded, optional, no NULLs 1.04 570.1±29.19µs? ?/sec1.00 549.2±5.53µs? ?/sec arrow_array_reader/BinaryViewArray/dictionary encoded, mandatory, no NULLs 1.00238.2±5.00µs? ?/sec1.06 252.5±15.70µs? ?/sec arrow_array_reader/BinaryViewArray/dictionary encoded, optional, half NULLs 1.05241.3±1.81µs? ?/sec1.00 229.9±2.92µs? ?/sec arrow_array_reader/BinaryViewArray/dictionary encoded, optional, no NULLs 1.00235.2±3.05µs? ?/sec1.09 255.7±3.29µs? ?/sec arrow_array_reader/BinaryViewArray/plain encoded, mandatory, no NULLs 1.00293.4±4.22µs? ?/sec1.00 292.7±4.54µs? ?/sec arrow_array_reader/BinaryViewArray/plain encoded, mandatory, no NULLs, short string1.00284.1±3.15µs? ?/sec1.07 305.1±3.08µs? ?/sec arrow_array_reader/BinaryViewArray/plain encoded, optional, half NULLs 1.07283.8±2.37µs? ?/sec1.00 266.2±2.03µs? ?/sec arrow_array_reader/BinaryViewArray/plain encoded, optional, no NULLs 1.00300.7±6.12µs? ?/sec1.00 302.0±1.53µs? ?/sec arrow_array_reader/FIXED_LEN_BYTE_ARRAY/Decimal128Array/byte_stream_split encoded, mandatory, no NULLs 1.05 1124.7±6.56µs? ?/sec1.00 1070.5±7.75µs? ?/sec arrow_array_reader/FIXED_LEN_BYTE_ARRAY/Decimal128Array/byte_stream_split encoded, optional, half NULLs1.04 976.8±11.17µs? ?/sec1.00 942.3±10.08µs? ?/sec arrow_array_reader/FIXED_LEN_BYTE_ARRAY/Decimal128Array/byte_stream_split encoded, optional, no NULLs 1.05 1136.5±18.59µs? ?/sec1.00 1083.9±15.25µs? ?/sec arrow_array_reader/FIXED_LEN_BYTE_ARRAY/Decimal128Array/plain encoded, mandatory, no NULLs 1.13453.9±9.08µs? ?/sec1.00 402.1±3.35µs? ?/sec arrow_array_reader/FIXED_LEN_BYTE_ARRAY/Decimal128Array/plain encoded, optional, half NULLs1.09 640.0±12.58µs? ?/sec1.00 588.2±6.46µs? ?/sec arrow_array_reader/FIXED_LEN_BYTE_ARRAY/Decimal128Array/plain encoded, optional, no NULLs 1.11458.7±3.46µs? ?/sec1.00 413.9±7.01µs? ?/sec arrow_array_reader/FIXED_LEN_BYTE_ARRAY/Float16Array/byte_stream_split encoded, mandatory, no NULLs1.00160.4±1.14µs? ?/sec1.26 202.2±1.23µs? ?/sec arrow_array_reader/FIXED_LEN_BYTE_ARRAY/Float16Array/byte_stream_split encoded, optional, half NULLs 1.00286.4±3.56µs? ?/sec1.11 316.5±2.25µs? ?/sec arrow_array_reader/FIXED_L
