This is an automated email from the ASF dual-hosted git repository.
alamb pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/arrow-rs.git
The following commit(s) were added to refs/heads/main by this push:
new 9d3a4d9f6f chore: add benchmark for row filters with LIMIT
short-circuit (#9767)
9d3a4d9f6f is described below
commit 9d3a4d9f6f91f421ec92641ac43b1abac90a0238
Author: Huaijin <[email protected]>
AuthorDate: Sun Apr 19 21:27:17 2026 +0800
chore: add benchmark for row filters with LIMIT short-circuit (#9767)
# Which issue does this PR close?
<!--
We generally require a GitHub issue to be filed for all bug fixes and
enhancements and this helps us generate change logs for our releases.
You can link an issue to this PR using the GitHub syntax.
-->
- part of #9766
# Rationale for this change
<!--
Why are you proposing this change? If this is already explained clearly
in the issue then this section is not needed.
Explaining clearly why changes are proposed helps reviewers understand
your changes and offer better suggestions for fixes.
-->
# What changes are included in this PR?
<!--
There is no need to duplicate the description in the issue here but it
is sometimes worth providing a summary of the individual changes in this
PR.
-->
# Are these changes tested?
<!--
We typically require tests for all PRs in order to:
1. Prevent the code from being accidentally broken by subsequent changes
2. Serve as another way to document the expected behavior of the code
If tests are not included in your PR, please explain why (for example,
are they covered by existing tests)?
-->
# Are there any user-facing changes?
<!--
If there are user-facing changes then we may require documentation to be
updated before approving the PR.
If there are any breaking changes to public APIs, please call them out.
-->
---
parquet/benches/arrow_reader_row_filter.rs | 117 ++++++++++++++++++++++++++++-
1 file changed, 114 insertions(+), 3 deletions(-)
diff --git a/parquet/benches/arrow_reader_row_filter.rs
b/parquet/benches/arrow_reader_row_filter.rs
index 331f5617ca..2b5a09eebc 100644
--- a/parquet/benches/arrow_reader_row_filter.rs
+++ b/parquet/benches/arrow_reader_row_filter.rs
@@ -180,10 +180,16 @@ fn create_record_batch(size: usize) -> RecordBatch {
RecordBatch::try_new(schema, arrays).unwrap()
}
+/// Total number of rows.
+const TOTAL_ROWS: usize = 500_000;
+
+/// Maximum rows per row group.
+const ROW_GROUP_SIZE: usize = 100_000;
+
/// Writes the RecordBatch to an in memory buffer, returning the buffer
fn write_parquet_file() -> Vec<u8> {
- let batch = create_record_batch(100_000);
- println!("Batch created with {} rows", 100_000);
+ let batch = create_record_batch(TOTAL_ROWS);
+ println!("Batch created with {TOTAL_ROWS} rows, row group size =
{ROW_GROUP_SIZE}");
println!(
"First 100 rows:\n{}",
pretty_format_batches(&[batch.clone().slice(0, 100)]).unwrap()
@@ -191,6 +197,7 @@ fn write_parquet_file() -> Vec<u8> {
let schema = batch.schema();
let props = WriterProperties::builder()
.set_compression(Compression::SNAPPY)
+ .set_max_row_group_row_count(Some(ROW_GROUP_SIZE))
.build();
let mut buffer = vec![];
{
@@ -522,6 +529,28 @@ async fn benchmark_async_reader(
}
}
+/// Like [`benchmark_async_reader`] but also threads `with_limit(limit)` into
+/// the stream builder. Used by the `LIMIT` benchmark below.
+async fn benchmark_async_reader_with_limit(
+ reader: InMemoryReader,
+ projection_mask: ProjectionMask,
+ row_filter: RowFilter,
+ limit: usize,
+) {
+ let mut stream = ParquetRecordBatchStreamBuilder::new(reader)
+ .await
+ .unwrap()
+ .with_batch_size(8192)
+ .with_projection(projection_mask)
+ .with_row_filter(row_filter)
+ .with_limit(limit)
+ .build()
+ .unwrap();
+ while let Some(b) = stream.next().await {
+ b.unwrap(); // consume the batches, no buffering
+ }
+}
+
/// Use sync API
fn benchmark_sync_reader(
reader: InMemoryReader,
@@ -586,5 +615,87 @@ impl AsyncFileReader for InMemoryReader {
}
}
-criterion_group!(benches, benchmark_filters_and_projections,);
+/// Benchmark filters with `LIMIT` short-circuit (`with_limit(N)`)
+///
+/// `PointLookup` is excluded because the filter has only 1 match in the
+/// whole file; `LIMIT 10` is not binding.
+fn benchmark_filters_with_limit(c: &mut Criterion) {
+ const LIMIT: usize = 10;
+
+ let parquet_file = Bytes::from(write_parquet_file());
+ let filter_types = vec![
+ FilterType::SelectiveUnclustered,
+ FilterType::ModeratelySelectiveClustered,
+ FilterType::ModeratelySelectiveUnclustered,
+ FilterType::UnselectiveUnclustered,
+ FilterType::UnselectiveClustered,
+ FilterType::Utf8ViewNonEmpty,
+ FilterType::Composite,
+ ];
+ let projection_cases = vec![
+ ProjectionCase::AllColumns,
+ ProjectionCase::ExcludeFilterColumn,
+ ];
+ let all_indices = vec![0, 1, 2, 3];
+
+ let rt = tokio::runtime::Builder::new_multi_thread()
+ .enable_all()
+ .build()
+ .unwrap();
+
+ let mut group = c.benchmark_group("arrow_reader_row_filter_limit");
+
+ for filter_type in filter_types {
+ for proj_case in &projection_cases {
+ let filter_col = filter_type.filter_projection().to_vec();
+ let output_projection: Vec<usize> = match proj_case {
+ ProjectionCase::AllColumns => all_indices.clone(),
+ ProjectionCase::ExcludeFilterColumn => all_indices
+ .iter()
+ .copied()
+ .filter(|i| !filter_col.contains(i))
+ .collect(),
+ };
+
+ let reader = InMemoryReader::try_new(&parquet_file).unwrap();
+ let metadata = Arc::clone(reader.metadata());
+ let schema_descr = metadata.file_metadata().schema_descr();
+ let projection_mask = ProjectionMask::roots(schema_descr,
output_projection);
+ let pred_mask = ProjectionMask::roots(schema_descr, filter_col);
+
+ let benchmark_name =
format!("{filter_type}/{proj_case}/limit{LIMIT}");
+
+ // async variant
+ let bench_id = BenchmarkId::new(benchmark_name.clone(), "async");
+ let rt_handle = rt.handle().clone();
+ let pred_mask_async = pred_mask.clone();
+ let projection_mask_async = projection_mask.clone();
+ let reader_async = reader.clone();
+ group.bench_function(bench_id, |b| {
+ b.iter(|| {
+ let reader = reader_async.clone();
+ let pred_mask = pred_mask_async.clone();
+ let projection_mask = projection_mask_async.clone();
+ // RowFilter and ArrowPredicateFn are not Clone — fresh
each iter.
+ let predicate = ArrowPredicateFn::new(pred_mask, move
|batch: RecordBatch| {
+ Ok(filter_type.filter_batch(&batch).unwrap())
+ });
+ let row_filter = RowFilter::new(vec![Box::new(predicate)]);
+ rt_handle.block_on(benchmark_async_reader_with_limit(
+ reader,
+ projection_mask,
+ row_filter,
+ LIMIT,
+ ));
+ });
+ });
+ }
+ }
+}
+
+criterion_group!(
+ benches,
+ benchmark_filters_and_projections,
+ benchmark_filters_with_limit,
+);
criterion_main!(benches);
