Skip to content

TEST: enable pushdown_filters and reorder_filters by default #18873

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 2 commits into
base: main
Choose a base branch
from
Draft

TEST: enable pushdown_filters and reorder_filters by default #18873

wants to merge 2 commits into from

Conversation

@alamb
Copy link
Contributor

@alamb alamb commented Nov 21, 2025 •
edited
Loading

( I am using this PR to test, I don't intend to merge it yet )

Which issue does this PR close?

Rationale for this change

We have made non trivial progress in filter representation in Parquet. Let's see where performance is now.

What changes are included in this PR?

Are these changes tested?

By CI tests

Are there any user-facing changes?

@github-actions github-actions bot added documentation Improvements or additions to documentation optimizer Optimizer rules core Core DataFusion crate sqllogictest SQL Logic Tests (.slt) common Related to common crate proto Related to proto crate datasource Changes to the datasource crate labels Nov 21, 2025
@alamb
Copy link
Contributor Author

alamb commented Nov 21, 2025

🤖 ./gh_compare_branch.sh Benchmark Script 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/pushdown_filters_test (8a06d63) to d65fb86 diff using: tpch_mem clickbench_partitioned clickbench_extended
Results will be posted here when complete

@alamb
Copy link
Contributor Author

alamb commented Nov 21, 2025

🤖: Benchmark completed

Details 

Comparing HEAD and alamb_pushdown_filters_test
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━┓
┃ Query        ┃        HEAD ┃ alamb_pushdown_filters_test ┃         Change ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━┩
│ QQuery 0     │  2704.76 ms │                  2693.83 ms │      no change │
│ QQuery 1     │  1211.14 ms │                  1278.46 ms │   1.06x slower │
│ QQuery 2     │  2313.54 ms │                  2408.79 ms │      no change │
│ QQuery 3     │  1159.95 ms │                  1156.61 ms │      no change │
│ QQuery 4     │  2336.83 ms │                  2351.37 ms │      no change │
│ QQuery 5     │ 28195.81 ms │                 27958.12 ms │      no change │
│ QQuery 6     │  4058.62 ms │                   108.14 ms │ +37.53x faster │
│ QQuery 7     │  3742.15 ms │                  3701.67 ms │      no change │
└──────────────┴─────────────┴─────────────────────────────┴────────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                          │ 45722.81ms │
│ Total Time (alamb_pushdown_filters_test)   │ 41656.99ms │
│ Average Time (HEAD)                        │  5715.35ms │
│ Average Time (alamb_pushdown_filters_test) │  5207.12ms │
│ Queries Faster                             │          1 │
│ Queries Slower                             │          1 │
│ Queries with No Change                     │          6 │
│ Queries with Failure                       │          0 │
└────────────────────────────────────────────┴────────────┘
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━┓
┃ Query        ┃        HEAD ┃ alamb_pushdown_filters_test ┃         Change ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━┩
│ QQuery 0     │     2.34 ms │                     2.61 ms │   1.12x slower │
│ QQuery 1     │    50.62 ms │                    51.87 ms │      no change │
│ QQuery 2     │   136.14 ms │                   136.47 ms │      no change │
│ QQuery 3     │   162.09 ms │                   160.21 ms │      no change │
│ QQuery 4     │  1098.61 ms │                  1095.76 ms │      no change │
│ QQuery 5     │  1477.67 ms │                  1491.19 ms │      no change │
│ QQuery 6     │     2.17 ms │                     2.16 ms │      no change │
│ QQuery 7     │    54.19 ms │                    68.48 ms │   1.26x slower │
│ QQuery 8     │  1444.55 ms │                  1422.06 ms │      no change │
│ QQuery 9     │  1890.35 ms │                  1852.51 ms │      no change │
│ QQuery 10    │   363.24 ms │                   484.80 ms │   1.33x slower │
│ QQuery 11    │   427.02 ms │                   532.87 ms │   1.25x slower │
│ QQuery 12    │  1340.34 ms │                  1545.03 ms │   1.15x slower │
│ QQuery 13    │  2077.33 ms │                  2280.93 ms │   1.10x slower │
│ QQuery 14    │  1274.21 ms │                  1448.68 ms │   1.14x slower │
│ QQuery 15    │  1277.86 ms │                  1261.35 ms │      no change │
│ QQuery 16    │  2677.07 ms │                  2704.20 ms │      no change │
│ QQuery 17    │  2662.55 ms │                  2692.03 ms │      no change │
│ QQuery 18    │  5448.25 ms │                  4951.08 ms │  +1.10x faster │
│ QQuery 19    │   130.47 ms │                   143.08 ms │   1.10x slower │
│ QQuery 20    │  2003.32 ms │                  1854.73 ms │  +1.08x faster │
│ QQuery 21    │  2355.36 ms │                  2330.38 ms │      no change │
│ QQuery 22    │  3991.87 ms │                  3944.03 ms │      no change │
│ QQuery 23    │ 16491.78 ms │                  1080.71 ms │ +15.26x faster │
│ QQuery 24    │   215.96 ms │                   253.74 ms │   1.17x slower │
│ QQuery 25    │   489.11 ms │                   617.52 ms │   1.26x slower │
│ QQuery 26    │   228.11 ms │                   324.39 ms │   1.42x slower │
│ QQuery 27    │  2868.16 ms │                  2929.86 ms │      no change │
│ QQuery 28    │ 23345.61 ms │                 23660.66 ms │      no change │
│ QQuery 29    │   975.88 ms │                   994.78 ms │      no change │
│ QQuery 30    │  1333.55 ms │                  1354.66 ms │      no change │
│ QQuery 31    │  1416.85 ms │                  1351.01 ms │      no change │
│ QQuery 32    │  5442.42 ms │                  4533.41 ms │  +1.20x faster │
│ QQuery 33    │  6011.79 ms │                  5754.30 ms │      no change │
│ QQuery 34    │  6139.89 ms │                  5858.52 ms │      no change │
│ QQuery 35    │  1950.34 ms │                  1865.06 ms │      no change │
│ QQuery 36    │   121.36 ms │                    25.69 ms │  +4.72x faster │
│ QQuery 37    │    51.21 ms │                    25.18 ms │  +2.03x faster │
│ QQuery 38    │   120.24 ms │                    25.57 ms │  +4.70x faster │
│ QQuery 39    │   194.03 ms │                    25.93 ms │  +7.48x faster │
│ QQuery 40    │    43.31 ms │                    26.00 ms │  +1.67x faster │
│ QQuery 41    │    40.36 ms │                    25.77 ms │  +1.57x faster │
│ QQuery 42    │    31.97 ms │                    25.25 ms │  +1.27x faster │
└──────────────┴─────────────┴─────────────────────────────┴────────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                          │ 99859.52ms │
│ Total Time (alamb_pushdown_filters_test)   │ 83214.54ms │
│ Average Time (HEAD)                        │  2322.31ms │
│ Average Time (alamb_pushdown_filters_test) │  1935.22ms │
│ Queries Faster                             │         11 │
│ Queries Slower                             │         11 │
│ Queries with No Change                     │         21 │
│ Queries with Failure                       │          0 │
└────────────────────────────────────────────┴────────────┘
--------------------
Benchmark tpch_mem_sf1.json
--------------------
┏━━━━━━━━━━━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query        ┃      HEAD ┃ alamb_pushdown_filters_test ┃        Change ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1     │ 131.35 ms │                   139.34 ms │  1.06x slower │
│ QQuery 2     │  29.56 ms │                    28.82 ms │     no change │
│ QQuery 3     │  34.98 ms │                    39.69 ms │  1.13x slower │
│ QQuery 4     │  29.36 ms │                    29.42 ms │     no change │
│ QQuery 5     │  87.73 ms │                    88.77 ms │     no change │
│ QQuery 6     │  19.59 ms │                    19.82 ms │     no change │
│ QQuery 7     │ 217.18 ms │                   221.16 ms │     no change │
│ QQuery 8     │  32.12 ms │                    34.75 ms │  1.08x slower │
│ QQuery 9     │ 103.41 ms │                   107.77 ms │     no change │
│ QQuery 10    │  64.14 ms │                    64.51 ms │     no change │
│ QQuery 11    │  19.10 ms │                    17.62 ms │ +1.08x faster │
│ QQuery 12    │  52.10 ms │                    51.44 ms │     no change │
│ QQuery 13    │  47.55 ms │                    46.22 ms │     no change │
│ QQuery 14    │  13.94 ms │                    13.57 ms │     no change │
│ QQuery 15    │  25.01 ms │                    24.55 ms │     no change │
│ QQuery 16    │  25.81 ms │                    25.46 ms │     no change │
│ QQuery 17    │ 148.10 ms │                   148.29 ms │     no change │
│ QQuery 18    │ 282.55 ms │                   280.77 ms │     no change │
│ QQuery 19    │  49.21 ms │                    38.45 ms │ +1.28x faster │
│ QQuery 20    │  50.67 ms │                    49.56 ms │     no change │
│ QQuery 21    │ 338.10 ms │                   320.93 ms │ +1.05x faster │
│ QQuery 22    │  17.73 ms │                    17.73 ms │     no change │
└──────────────┴───────────┴─────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                          │ 1819.30ms │
│ Total Time (alamb_pushdown_filters_test)   │ 1808.63ms │
│ Average Time (HEAD)                        │   82.70ms │
│ Average Time (alamb_pushdown_filters_test) │   82.21ms │
│ Queries Faster                             │         3 │
│ Queries Slower                             │         3 │
│ Queries with No Change                     │        16 │
│ Queries with Failure                       │         0 │
└────────────────────────────────────────────┴───────────┘

@alamb alamb changed the title TEST: enable pushdown_filters by default TEST: enable pushdown_filters and reorder_filters by default Nov 21, 2025
@alamb alamb mentioned this pull request Nov 21, 2025
Draft
@alamb
Copy link
Contributor Author

alamb commented Nov 21, 2025

I am also testing with just filter_pushdown on:

I am going to focus my efforts on profiling these queries which seem to have gotten the most slower:

│ QQuery 24    │   215.96 ms │                   253.74 ms │   1.17x slower │
│ QQuery 25    │   489.11 ms │                   617.52 ms │   1.26x slower │
│ QQuery 26    │   228.11 ms │                   324.39 ms │   1.42x slower │

Here is the query:

set datafusion.execution.parquet.binary_as_string = true
SELECT "SearchPhrase" FROM hits WHERE "SearchPhrase" <> '' ORDER BY "SearchPhrase" LIMIT 10;

Basically my next steps are to profile these queries and see what is slower (and if it is related to filter representation, I will go focus on apache/arrow-rs#8902)

@Dandandan
Copy link
Contributor

Dandandan commented Nov 22, 2025

I am also testing with just filter_pushdown on:

I am going to focus my efforts on profiling these queries which seem to have gotten the most slower:

│ QQuery 24    │   215.96 ms │                   253.74 ms │   1.17x slower │
│ QQuery 25    │   489.11 ms │                   617.52 ms │   1.26x slower │
│ QQuery 26    │   228.11 ms │                   324.39 ms │   1.42x slower │

Here is the query:

set datafusion.execution.parquet.binary_as_string = true
SELECT "SearchPhrase" FROM hits WHERE "SearchPhrase" <> '' ORDER BY "SearchPhrase" LIMIT 10;

Basically my next steps are to profile these queries and see what is slower (and if it is related to filter representation, I will go focus on apache/arrow-rs#8902)

Looks like we are very close!

FYI, there a couple more slow than query 24:

│ QQuery 26    │   228.11 ms │                   324.39 ms │   1.42x slower │
│ QQuery 10    │   363.24 ms │                   484.80 ms │   1.33x slower │
│ QQuery 7     │    54.19 ms │                    68.48 ms │   1.26x slower │
│ QQuery 25    │   489.11 ms │                   617.52 ms │   1.26x slower │
│ QQuery 11    │   427.02 ms │                   532.87 ms │   1.25x slower │

@alamb
Copy link
Contributor Author

alamb commented Nov 23, 2025

I did some more analysis:

The idea is to isolate why filter pushdown is slowing down clickbench q24

See more details here #18873

This is after upgrading to arrow 57.1.0

The only difference in the two binaries is if filter pushdown is on by default:

-rwxr-xr-x@ 1 andrewlamb  staff  81331152 Nov 23 07:31 datafusion-cli-alamb_upgrade_arrow_57.1.0
-rwxr-xr-x@ 1 andrewlamb  staff  81331152 Nov 22 07:57 datafusion-cli-almab_pushdown_no_reorder

Using hits partitioned dataset

ln -s ~/Software/datafusion/benchmarks/data/hits_partitioned ./hits

Here is q24.sql

set datafusion.execution.parquet.binary_as_string = true;

-- turn on pushdown (is hard coded)
-- set datafusion.execution.parquet.pushdown_filters = true;

SELECT "SearchPhrase" FROM hits WHERE "SearchPhrase" <> '' ORDER BY "SearchPhrase" LIMIT 10;
SELECT "SearchPhrase" FROM hits WHERE "SearchPhrase" <> '' ORDER BY "SearchPhrase" LIMIT 10;
SELECT "SearchPhrase" FROM hits WHERE "SearchPhrase" <> '' ORDER BY "SearchPhrase" LIMIT 10;
SELECT "SearchPhrase" FROM hits WHERE "SearchPhrase" <> '' ORDER BY "SearchPhrase" LIMIT 10;
SELECT "SearchPhrase" FROM hits WHERE "SearchPhrase" <> '' ORDER BY "SearchPhrase" LIMIT 10;
SELECT "SearchPhrase" FROM hits WHERE "SearchPhrase" <> '' ORDER BY "SearchPhrase" LIMIT 10;
SELECT "SearchPhrase" FROM hits WHERE "SearchPhrase" <> '' ORDER BY "SearchPhrase" LIMIT 10;
SELECT "SearchPhrase" FROM hits WHERE "SearchPhrase" <> '' ORDER BY "SearchPhrase" LIMIT 10;
SELECT "SearchPhrase" FROM hits WHERE "SearchPhrase" <> '' ORDER BY "SearchPhrase" LIMIT 10;
SELECT "SearchPhrase" FROM hits WHERE "SearchPhrase" <> '' ORDER BY "SearchPhrase" LIMIT 10;

You can see the pushdown is slightly slower

./datafusion-cli-almab_pushdown_no_reorder -f q24.sql  | grep Elapsed
Elapsed 0.000 seconds.
Elapsed 0.183 seconds.
Elapsed 0.154 seconds.
Elapsed 0.155 seconds.
Elapsed 0.153 seconds.
Elapsed 0.154 seconds.
Elapsed 0.154 seconds.
Elapsed 0.150 seconds.
Elapsed 0.154 seconds.
Elapsed 0.156 seconds.
Elapsed 0.152 seconds.
./datafusion-cli-alamb_upgrade_arrow_57.1.0  -f q24.sql  | grep Elapsed
Elapsed 0.002 seconds.
Elapsed 0.164 seconds.
Elapsed 0.137 seconds.
Elapsed 0.137 seconds.
Elapsed 0.133 seconds.
Elapsed 0.132 seconds.
Elapsed 0.135 seconds.
Elapsed 0.131 seconds.
Elapsed 0.137 seconds.
Elapsed 0.137 seconds.
Elapsed 0.133 seconds.

So let's profile what the pushdown one is doing

Screenshot 2025-11-23 at 7 40 41 AM

So more than 5% of the time is being spent converting filters back and forth.

Thus, this gives me more motivation to keep working on

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

common Related to common crate core Core DataFusion crate datasource Changes to the datasource crate documentation Improvements or additions to documentation optimizer Optimizer rules proto Related to proto crate sqllogictest SQL Logic Tests (.slt)

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

Enable parquet filter pushdown (filter_pushdown) by default

2 participants

@alamb @Dandandan