Skip to content

Refactor filter functions as methods on FilterPredicate struct #7771

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Draft
wants to merge 1 commit into
base: main
Choose a base branch
from
Draft

Refactor filter functions as methods on FilterPredicate struct #7771

wants to merge 1 commit into from

Conversation

alamb
Copy link
Contributor

@alamb alamb commented Jun 25, 2025

(This is a draft for now until I have a PR that shows how it works)

Which issue does this PR close?

Rationale for this change

I would like to reuse some part of the existing filtering logic for #7762. However, it is currently split into a bunch of free functions that all take a FilterPredicate argument. By making them methods on FilterPredicate I will be able to reuse the code

What changes are included in this PR?

  1. Move filter_* methods into functions on FilterPredicate (note all the methods are still private so this doesn't change the public API)

Are these changes tested?

Existing CI covers them functionally

I will run performance tests to ensure we don't mess up performance

Are there any user-facing changes?

There are no intended user facing changes

@github-actions github-actions bot added the arrow Changes to the arrow crate label Jun 25, 2025
@alamb alamb changed the title Refactor coalesce kernel to use filter builder Refactor filter functions as methods on FilterPredicate struct Jun 25, 2025
@alamb
Copy link
Contributor Author

alamb commented Jun 25, 2025

🤖 ./gh_compare_arrow.sh Benchmark Script Running
Linux aal-dev 6.11.0-1015-gcp #15~24.04.1-Ubuntu SMP Thu Apr 24 20:41:05 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux
Comparing alamb/filter_predicate (0b70ea6) to b6240b3 diff
BENCH_NAME=filter_kernels
BENCH_COMMAND=cargo bench --features=arrow,async,test_common,experimental --bench filter_kernels
BENCH_FILTER=
BENCH_BRANCH_NAME=alamb_filter_predicate
Results will be posted here when complete

alamb
alamb commented Jun 25, 2025
View reviewed changes
arrow-select/src/filter.rs
@@ -180,25 +180,15 @@ pub fn prep_null_mask_filter(filter: &BooleanArray) -> BooleanArray {
pub fn filter(values: &dyn Array, predicate: &BooleanArray) -> Result<ArrayRef, ArrowError> {
let mut filter_builder = FilterBuilder::new(predicate);

if multiple_arrays(values.data_type()) {
if FilterBuilder::multiple_arrays(values.data_type()) {
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The diff makes this PR look like a much larger change than it really is -- I found using whitespace blind diff was easier to see the changes: https://github.com/apache/arrow-rs/pull/7771/files?w=1

@alamb
Copy link
Contributor Author

alamb commented Jun 25, 2025

🤖: Benchmark completed

Details 

group                                                                         alamb_filter_predicate                 main
-----                                                                         ----------------------                 ----
filter context decimal128 (kept 1/2)                                          1.00     39.9±2.68µs        ? ?/sec    1.07     42.5±4.51µs        ? ?/sec
filter context decimal128 high selectivity (kept 1023/1024)                   1.01     51.6±1.18µs        ? ?/sec    1.00     51.0±1.23µs        ? ?/sec
filter context decimal128 low selectivity (kept 1/1024)                       1.04    261.4±0.35ns        ? ?/sec    1.00    252.4±0.31ns        ? ?/sec
filter context f32 (kept 1/2)                                                 1.00     69.4±0.12µs        ? ?/sec    1.41     98.1±0.21µs        ? ?/sec
filter context f32 high selectivity (kept 1023/1024)                          1.04     13.8±0.55µs        ? ?/sec    1.00     13.2±0.65µs        ? ?/sec
filter context f32 low selectivity (kept 1/1024)                              1.00    448.0±0.61ns        ? ?/sec    1.10    492.6±0.57ns        ? ?/sec
filter context fsb with value length 20 (kept 1/2)                            1.00     42.5±0.09µs        ? ?/sec    1.67     70.9±0.14µs        ? ?/sec
filter context fsb with value length 20 high selectivity (kept 1023/1024)     1.00     42.5±0.13µs        ? ?/sec    1.67     70.9±0.09µs        ? ?/sec
filter context fsb with value length 20 low selectivity (kept 1/1024)         1.00     42.4±0.10µs        ? ?/sec    1.67     70.9±0.14µs        ? ?/sec
filter context fsb with value length 5 (kept 1/2)                             1.00     42.4±0.11µs        ? ?/sec    1.67     71.0±0.12µs        ? ?/sec
filter context fsb with value length 5 high selectivity (kept 1023/1024)      1.00     42.4±0.07µs        ? ?/sec    1.67     70.9±0.10µs        ? ?/sec
filter context fsb with value length 5 low selectivity (kept 1/1024)          1.00     42.4±0.08µs        ? ?/sec    1.67     71.0±0.14µs        ? ?/sec
filter context fsb with value length 50 (kept 1/2)                            1.00     42.4±0.10µs        ? ?/sec    1.67     71.0±0.15µs        ? ?/sec
filter context fsb with value length 50 high selectivity (kept 1023/1024)     1.00     42.4±0.08µs        ? ?/sec    1.67     70.9±0.10µs        ? ?/sec
filter context fsb with value length 50 low selectivity (kept 1/1024)         1.00     42.4±0.08µs        ? ?/sec    1.67     71.0±0.09µs        ? ?/sec
filter context i32 (kept 1/2)                                                 1.01     22.9±0.06µs        ? ?/sec    1.00     22.7±0.05µs        ? ?/sec
filter context i32 high selectivity (kept 1023/1024)                          1.04      6.5±0.46µs        ? ?/sec    1.00      6.3±0.39µs        ? ?/sec
filter context i32 low selectivity (kept 1/1024)                              1.02    262.5±0.60ns        ? ?/sec    1.00    257.1±0.32ns        ? ?/sec
filter context i32 w NULLs (kept 1/2)                                         1.00     65.7±0.10µs        ? ?/sec    1.43     94.2±0.21µs        ? ?/sec
filter context i32 w NULLs high selectivity (kept 1023/1024)                  1.02     13.4±0.47µs        ? ?/sec    1.00     13.1±0.40µs        ? ?/sec
filter context i32 w NULLs low selectivity (kept 1/1024)                      1.00    455.5±0.72ns        ? ?/sec    1.06    483.5±0.58ns        ? ?/sec
filter context mixed string view (kept 1/2)                                   1.00     90.1±7.57µs        ? ?/sec    1.35    121.5±6.20µs        ? ?/sec
filter context mixed string view high selectivity (kept 1023/1024)            1.00     60.2±0.55µs        ? ?/sec    1.00     60.0±0.59µs        ? ?/sec
filter context mixed string view low selectivity (kept 1/1024)                1.00    634.7±0.60ns        ? ?/sec    1.09    689.8±1.07ns        ? ?/sec
filter context short string view (kept 1/2)                                   1.00     87.2±7.12µs        ? ?/sec    1.40    122.5±6.49µs        ? ?/sec
filter context short string view high selectivity (kept 1023/1024)            1.01     58.6±0.76µs        ? ?/sec    1.00     58.2±1.06µs        ? ?/sec
filter context short string view low selectivity (kept 1/1024)                1.00    448.4±0.63ns        ? ?/sec    1.12    501.2±0.76ns        ? ?/sec
filter context string (kept 1/2)                                              1.00    550.9±4.02µs        ? ?/sec    1.12   614.6±13.48µs        ? ?/sec
filter context string dictionary (kept 1/2)                                   1.00     23.3±0.05µs        ? ?/sec    1.00     23.3±0.06µs        ? ?/sec
filter context string dictionary high selectivity (kept 1023/1024)            1.00      7.1±0.35µs        ? ?/sec    1.06      7.5±0.36µs        ? ?/sec
filter context string dictionary low selectivity (kept 1/1024)                1.00    847.8±1.87ns        ? ?/sec    1.01    852.1±4.13ns        ? ?/sec
filter context string dictionary w NULLs (kept 1/2)                           1.00     66.6±0.15µs        ? ?/sec    1.43     95.0±0.33µs        ? ?/sec
filter context string dictionary w NULLs high selectivity (kept 1023/1024)    1.07     14.7±0.48µs        ? ?/sec    1.00     13.8±0.37µs        ? ?/sec
filter context string dictionary w NULLs low selectivity (kept 1/1024)        1.00   1062.8±1.95ns        ? ?/sec    1.04   1106.0±3.81ns        ? ?/sec
filter context string high selectivity (kept 1023/1024)                       1.00   627.0±16.33µs        ? ?/sec    1.06   662.8±36.08µs        ? ?/sec
filter context string low selectivity (kept 1/1024)                           1.00   1089.5±6.02ns        ? ?/sec    1.04   1135.2±1.61ns        ? ?/sec
filter context u8 (kept 1/2)                                                  1.00     18.9±0.04µs        ? ?/sec    1.00     18.9±0.05µs        ? ?/sec
filter context u8 high selectivity (kept 1023/1024)                           1.01   1814.5±8.56ns        ? ?/sec    1.00  1804.2±11.39ns        ? ?/sec
filter context u8 low selectivity (kept 1/1024)                               1.01    250.5±0.38ns        ? ?/sec    1.00    247.5±0.92ns        ? ?/sec
filter context u8 w NULLs (kept 1/2)                                          1.00     61.6±0.11µs        ? ?/sec    1.46     90.2±0.22µs        ? ?/sec
filter context u8 w NULLs high selectivity (kept 1023/1024)                   1.06      8.6±0.01µs        ? ?/sec    1.00      8.2±0.06µs        ? ?/sec
filter context u8 w NULLs low selectivity (kept 1/1024)                       1.00    437.3±3.15ns        ? ?/sec    1.30    569.5±0.58ns        ? ?/sec
filter decimal128 (kept 1/2)                                                  1.00     94.7±0.35µs        ? ?/sec    1.03     97.3±0.34µs        ? ?/sec
filter decimal128 high selectivity (kept 1023/1024)                           1.00     52.3±1.98µs        ? ?/sec    1.00     52.3±1.55µs        ? ?/sec
filter decimal128 low selectivity (kept 1/1024)                               1.23      3.0±0.01µs        ? ?/sec    1.00      2.5±0.01µs        ? ?/sec
filter f32 (kept 1/2)                                                         1.18    230.6±0.32µs        ? ?/sec    1.00    195.9±0.58µs        ? ?/sec
filter fsb with value length 20 (kept 1/2)                                    1.00    132.9±0.30µs        ? ?/sec    1.14    151.5±0.38µs        ? ?/sec
filter fsb with value length 20 high selectivity (kept 1023/1024)             1.00     69.5±2.02µs        ? ?/sec    1.05     73.0±1.71µs        ? ?/sec
filter fsb with value length 20 low selectivity (kept 1/1024)                 1.24      3.2±0.01µs        ? ?/sec    1.00      2.6±0.00µs        ? ?/sec
filter fsb with value length 5 (kept 1/2)                                     1.00    131.2±0.23µs        ? ?/sec    1.17    152.9±0.28µs        ? ?/sec
filter fsb with value length 5 high selectivity (kept 1023/1024)              1.00     11.1±0.63µs        ? ?/sec    1.02     11.3±0.46µs        ? ?/sec
filter fsb with value length 5 low selectivity (kept 1/1024)                  1.23      3.1±0.01µs        ? ?/sec    1.00      2.5±0.01µs        ? ?/sec
filter fsb with value length 50 (kept 1/2)                                    1.00    183.0±7.00µs        ? ?/sec    1.06    194.0±9.68µs        ? ?/sec
filter fsb with value length 50 high selectivity (kept 1023/1024)             1.00   213.1±10.85µs        ? ?/sec    1.07    228.5±4.31µs        ? ?/sec
filter fsb with value length 50 low selectivity (kept 1/1024)                 1.22      3.1±0.01µs        ? ?/sec    1.00      2.6±0.00µs        ? ?/sec
filter i32 (kept 1/2)                                                         1.01     92.3±0.13µs        ? ?/sec    1.00     91.5±0.33µs        ? ?/sec
filter i32 high selectivity (kept 1023/1024)                                  1.00      8.4±0.34µs        ? ?/sec    1.02      8.5±0.21µs        ? ?/sec
filter i32 low selectivity (kept 1/1024)                                      1.25      3.1±0.00µs        ? ?/sec    1.00      2.5±0.01µs        ? ?/sec
filter optimize (kept 1/2)                                                    1.02     84.1±0.21µs        ? ?/sec    1.00     82.5±0.31µs        ? ?/sec
filter optimize high selectivity (kept 1023/1024)                             1.00      2.7±0.01µs        ? ?/sec    1.16      3.1±0.01µs        ? ?/sec
filter optimize low selectivity (kept 1/1024)                                 1.31      2.8±0.00µs        ? ?/sec    1.00      2.1±0.01µs        ? ?/sec
filter run array (kept 1/2)                                                   1.00    442.5±0.85µs        ? ?/sec    1.06    469.7±0.98µs        ? ?/sec
filter run array high selectivity (kept 1023/1024)                            1.00    397.6±1.03µs        ? ?/sec    1.13    448.1±1.42µs        ? ?/sec
filter run array low selectivity (kept 1/1024)                                1.00    333.5±0.76µs        ? ?/sec    1.00    332.9±0.91µs        ? ?/sec
filter single record batch                                                    1.02     93.3±0.25µs        ? ?/sec    1.00     91.6±0.19µs        ? ?/sec
filter u8 (kept 1/2)                                                          1.00     91.9±0.10µs        ? ?/sec    1.04     95.4±0.21µs        ? ?/sec
filter u8 high selectivity (kept 1023/1024)                                   1.00      3.8±0.01µs        ? ?/sec    1.09      4.1±0.01µs        ? ?/sec
filter u8 low selectivity (kept 1/1024)                                       1.24      3.0±0.01µs        ? ?/sec    1.00      2.5±0.01µs        ? ?/sec

@alamb
Copy link
Contributor Author

alamb commented Jun 25, 2025

🤖: Benchmark completed

Wow those are some noisy benchmarks 🤔

@alamb
Copy link
Contributor Author

alamb commented Jun 25, 2025

🤖 ./gh_compare_arrow.sh Benchmark Script Running
Linux aal-dev 6.11.0-1015-gcp #15~24.04.1-Ubuntu SMP Thu Apr 24 20:41:05 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux
Comparing alamb/filter_predicate (0b70ea6) to b6240b3 diff
BENCH_NAME=filter_kernels
BENCH_COMMAND=cargo bench --features=arrow,async,test_common,experimental --bench filter_kernels
BENCH_FILTER=
BENCH_BRANCH_NAME=alamb_filter_predicate
Results will be posted here when complete

@alamb
Copy link
Contributor Author

alamb commented Jun 25, 2025

🤖: Benchmark completed

Details 

group                                                                         alamb_filter_predicate                 main
-----                                                                         ----------------------                 ----
filter context decimal128 (kept 1/2)                                          1.17     51.6±7.87µs        ? ?/sec    1.00     44.0±7.01µs        ? ?/sec
filter context decimal128 high selectivity (kept 1023/1024)                   1.00     50.5±0.51µs        ? ?/sec    1.01     51.0±1.32µs        ? ?/sec
filter context decimal128 low selectivity (kept 1/1024)                       1.00    249.6±0.41ns        ? ?/sec    1.02    254.1±0.36ns        ? ?/sec
filter context f32 (kept 1/2)                                                 1.00     69.4±0.12µs        ? ?/sec    1.41     98.0±0.18µs        ? ?/sec
filter context f32 high selectivity (kept 1023/1024)                          1.00     13.7±0.60µs        ? ?/sec    1.01     13.8±0.50µs        ? ?/sec
filter context f32 low selectivity (kept 1/1024)                              1.00    448.7±0.36ns        ? ?/sec    1.10    494.8±1.85ns        ? ?/sec
filter context fsb with value length 20 (kept 1/2)                            1.00     42.4±0.07µs        ? ?/sec    1.67     71.0±0.12µs        ? ?/sec
filter context fsb with value length 20 high selectivity (kept 1023/1024)     1.00     42.4±0.05µs        ? ?/sec    1.67     71.0±0.16µs        ? ?/sec
filter context fsb with value length 20 low selectivity (kept 1/1024)         1.00     42.4±0.06µs        ? ?/sec    1.67     71.0±0.16µs        ? ?/sec
filter context fsb with value length 5 (kept 1/2)                             1.00     42.4±0.06µs        ? ?/sec    1.67     71.0±0.15µs        ? ?/sec
filter context fsb with value length 5 high selectivity (kept 1023/1024)      1.00     42.4±0.10µs        ? ?/sec    1.68     71.1±0.55µs        ? ?/sec
filter context fsb with value length 5 low selectivity (kept 1/1024)          1.00     42.4±0.06µs        ? ?/sec    1.67     71.0±0.17µs        ? ?/sec
filter context fsb with value length 50 (kept 1/2)                            1.00     42.5±0.07µs        ? ?/sec    1.67     71.1±0.32µs        ? ?/sec
filter context fsb with value length 50 high selectivity (kept 1023/1024)     1.00     42.5±0.21µs        ? ?/sec    1.67     71.0±0.14µs        ? ?/sec
filter context fsb with value length 50 low selectivity (kept 1/1024)         1.00     42.4±0.07µs        ? ?/sec    1.67     71.0±0.10µs        ? ?/sec
filter context i32 (kept 1/2)                                                 1.00     22.8±0.05µs        ? ?/sec    1.00     22.8±0.06µs        ? ?/sec
filter context i32 high selectivity (kept 1023/1024)                          1.00      6.4±0.36µs        ? ?/sec    1.02      6.5±0.38µs        ? ?/sec
filter context i32 low selectivity (kept 1/1024)                              1.01    258.7±0.39ns        ? ?/sec    1.00    256.0±0.41ns        ? ?/sec
filter context i32 w NULLs (kept 1/2)                                         1.00     65.7±0.13µs        ? ?/sec    1.43     94.1±0.19µs        ? ?/sec
filter context i32 w NULLs high selectivity (kept 1023/1024)                  1.05     13.6±0.59µs        ? ?/sec    1.00     12.9±0.43µs        ? ?/sec
filter context i32 w NULLs low selectivity (kept 1/1024)                      1.00    458.4±0.73ns        ? ?/sec    1.05    482.2±0.59ns        ? ?/sec
filter context mixed string view (kept 1/2)                                   1.00     89.3±6.68µs        ? ?/sec    1.32    118.1±8.03µs        ? ?/sec
filter context mixed string view high selectivity (kept 1023/1024)            1.00     58.2±1.24µs        ? ?/sec    1.00     58.2±1.79µs        ? ?/sec
filter context mixed string view low selectivity (kept 1/1024)                1.00    629.4±0.78ns        ? ?/sec    1.08    679.1±2.28ns        ? ?/sec
filter context short string view (kept 1/2)                                   1.00     85.3±5.36µs        ? ?/sec    1.32    112.3±4.76µs        ? ?/sec
filter context short string view high selectivity (kept 1023/1024)            1.00     57.5±2.10µs        ? ?/sec    1.02     58.4±1.61µs        ? ?/sec
filter context short string view low selectivity (kept 1/1024)                1.00    447.3±0.47ns        ? ?/sec    1.09    488.2±1.06ns        ? ?/sec
filter context string (kept 1/2)                                              1.00   559.3±13.55µs        ? ?/sec    1.08   602.1±14.95µs        ? ?/sec
filter context string dictionary (kept 1/2)                                   1.00     23.2±0.04µs        ? ?/sec    1.00     23.3±0.04µs        ? ?/sec
filter context string dictionary high selectivity (kept 1023/1024)            1.00      7.1±0.57µs        ? ?/sec    1.05      7.5±0.30µs        ? ?/sec
filter context string dictionary low selectivity (kept 1/1024)                1.00    818.6±1.02ns        ? ?/sec    1.00    821.6±1.80ns        ? ?/sec
filter context string dictionary w NULLs (kept 1/2)                           1.00     66.6±0.11µs        ? ?/sec    1.43     95.0±0.22µs        ? ?/sec
filter context string dictionary w NULLs high selectivity (kept 1023/1024)    1.05     14.7±0.45µs        ? ?/sec    1.00     13.9±0.33µs        ? ?/sec
filter context string dictionary w NULLs low selectivity (kept 1/1024)        1.00   1031.3±1.18ns        ? ?/sec    1.04   1076.1±2.31ns        ? ?/sec
filter context string high selectivity (kept 1023/1024)                       1.00   638.6±18.95µs        ? ?/sec    1.04   665.2±19.95µs        ? ?/sec
filter context string low selectivity (kept 1/1024)                           1.00    880.9±2.19ns        ? ?/sec    1.39   1222.5±3.96ns        ? ?/sec
filter context u8 (kept 1/2)                                                  1.00     18.9±0.05µs        ? ?/sec    1.00     18.9±0.03µs        ? ?/sec
filter context u8 high selectivity (kept 1023/1024)                           1.00   1810.0±7.95ns        ? ?/sec    1.01  1833.0±11.09ns        ? ?/sec
filter context u8 low selectivity (kept 1/1024)                               1.00    246.4±0.38ns        ? ?/sec    1.01    247.7±0.40ns        ? ?/sec
filter context u8 w NULLs (kept 1/2)                                          1.00     61.7±0.14µs        ? ?/sec    1.46     90.2±0.19µs        ? ?/sec
filter context u8 w NULLs high selectivity (kept 1023/1024)                   1.06      8.7±0.03µs        ? ?/sec    1.00      8.2±0.02µs        ? ?/sec
filter context u8 w NULLs low selectivity (kept 1/1024)                       1.00    435.8±0.77ns        ? ?/sec    1.32    574.6±2.20ns        ? ?/sec
filter decimal128 (kept 1/2)                                                  1.00     94.8±0.46µs        ? ?/sec    1.03     97.5±0.46µs        ? ?/sec
filter decimal128 high selectivity (kept 1023/1024)                           1.00     52.5±1.12µs        ? ?/sec    1.01     53.3±0.55µs        ? ?/sec
filter decimal128 low selectivity (kept 1/1024)                               1.21      3.0±0.00µs        ? ?/sec    1.00      2.5±0.00µs        ? ?/sec
filter f32 (kept 1/2)                                                         1.18    230.5±0.37µs        ? ?/sec    1.00    195.8±0.24µs        ? ?/sec
filter fsb with value length 20 (kept 1/2)                                    1.00    133.6±0.69µs        ? ?/sec    1.14    151.7±0.79µs        ? ?/sec
filter fsb with value length 20 high selectivity (kept 1023/1024)             1.02     71.9±3.82µs        ? ?/sec    1.00     70.2±0.55µs        ? ?/sec
filter fsb with value length 20 low selectivity (kept 1/1024)                 1.24      3.2±0.01µs        ? ?/sec    1.00      2.5±0.01µs        ? ?/sec
filter fsb with value length 5 (kept 1/2)                                     1.00    131.4±0.23µs        ? ?/sec    1.16    152.8±0.33µs        ? ?/sec
filter fsb with value length 5 high selectivity (kept 1023/1024)              1.00     11.0±0.90µs        ? ?/sec    1.01     11.2±0.58µs        ? ?/sec
filter fsb with value length 5 low selectivity (kept 1/1024)                  1.24      3.1±0.00µs        ? ?/sec    1.00      2.5±0.00µs        ? ?/sec
filter fsb with value length 50 (kept 1/2)                                    1.00    185.5±5.08µs        ? ?/sec    1.03   191.7±13.00µs        ? ?/sec
filter fsb with value length 50 high selectivity (kept 1023/1024)             1.00    209.5±7.80µs        ? ?/sec    1.04    217.8±7.56µs        ? ?/sec
filter fsb with value length 50 low selectivity (kept 1/1024)                 1.21      3.1±0.01µs        ? ?/sec    1.00      2.6±0.02µs        ? ?/sec
filter i32 (kept 1/2)                                                         1.01     92.2±0.12µs        ? ?/sec    1.00     91.5±0.17µs        ? ?/sec
filter i32 high selectivity (kept 1023/1024)                                  1.00      8.7±0.40µs        ? ?/sec    1.01      8.8±0.38µs        ? ?/sec
filter i32 low selectivity (kept 1/1024)                                      1.24      3.1±0.01µs        ? ?/sec    1.00      2.5±0.01µs        ? ?/sec
filter optimize (kept 1/2)                                                    1.03     84.3±0.53µs        ? ?/sec    1.00     82.0±0.32µs        ? ?/sec
filter optimize high selectivity (kept 1023/1024)                             1.00      2.7±0.01µs        ? ?/sec    1.19      3.2±0.01µs        ? ?/sec
filter optimize low selectivity (kept 1/1024)                                 1.29      2.8±0.00µs        ? ?/sec    1.00      2.1±0.00µs        ? ?/sec
filter run array (kept 1/2)                                                   1.00    443.2±0.63µs        ? ?/sec    1.06    469.9±1.05µs        ? ?/sec
filter run array high selectivity (kept 1023/1024)                            1.00    397.4±2.01µs        ? ?/sec    1.12    445.8±1.54µs        ? ?/sec
filter run array low selectivity (kept 1/1024)                                1.00    333.3±0.90µs        ? ?/sec    1.00    333.4±1.35µs        ? ?/sec
filter single record batch                                                    1.02     93.4±0.17µs        ? ?/sec    1.00     92.0±0.46µs        ? ?/sec
filter u8 (kept 1/2)                                                          1.00     91.9±0.16µs        ? ?/sec    1.04     95.3±0.21µs        ? ?/sec
filter u8 high selectivity (kept 1023/1024)                                   1.00      3.7±0.01µs        ? ?/sec    1.07      4.0±0.01µs        ? ?/sec
filter u8 low selectivity (kept 1/1024)                                       1.22      3.0±0.00µs        ? ?/sec    1.00      2.5±0.01µs        ? ?/sec

@alamb
Copy link
Contributor Author

alamb commented Jun 25, 2025

My conclusion from the benchmark numbers is that this PR doesn't change performance, but there seems to be a lot of noise

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
arrow Changes to the arrow crate
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
1 participant
@alamb
pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy