Skip to content
Merged
Changes from 1 commit
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
200 changes: 139 additions & 61 deletions datafusion/physical-expr/benches/case_when.rs
Original file line number Diff line number Diff line change
Expand Up @@ -16,30 +16,29 @@
// under the License.

use arrow::array::builder::{Int32Builder, StringBuilder};
use arrow::datatypes::{DataType, Field, Schema};
use arrow::array::{Array, ArrayRef, Int32Array};
use arrow::datatypes::{Field, Schema};
use arrow::record_batch::RecordBatch;
use criterion::{black_box, criterion_group, criterion_main, Criterion};
use datafusion_common::ScalarValue;
use datafusion_expr::Operator;
use datafusion_physical_expr::expressions::{BinaryExpr, CaseExpr, Column, Literal};
use datafusion_physical_expr::expressions::{case, col, lit, BinaryExpr};
use datafusion_physical_expr_common::physical_expr::PhysicalExpr;
use std::sync::Arc;

fn make_col(name: &str, index: usize) -> Arc<dyn PhysicalExpr> {
Arc::new(Column::new(name, index))
fn make_x_cmp_y(
x: &Arc<dyn PhysicalExpr>,
op: Operator,
y: i32,
) -> Arc<dyn PhysicalExpr> {
Arc::new(BinaryExpr::new(Arc::clone(x), op, lit(y)))
}

fn make_lit_i32(n: i32) -> Arc<dyn PhysicalExpr> {
Arc::new(Literal::new(ScalarValue::Int32(Some(n))))
}

fn criterion_benchmark(c: &mut Criterion) {
// create input data
fn make_batch(row_count: usize, column_count: usize) -> RecordBatch {

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

is it worth a comment saying that this column could is always 3 or more?

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

I'll add a description and assertion

let mut c1 = Int32Builder::new();
let mut c2 = StringBuilder::new();
let mut c3 = StringBuilder::new();
for i in 0..1000 {
c1.append_value(i);
for i in 0..row_count {
c1.append_value(i as i32);
if i % 7 == 0 {
c2.append_null();
} else {
Expand All @@ -54,69 +53,148 @@ fn criterion_benchmark(c: &mut Criterion) {
let c1 = Arc::new(c1.finish());
let c2 = Arc::new(c2.finish());
let c3 = Arc::new(c3.finish());
let schema = Schema::new(vec![
Field::new("c1", DataType::Int32, true),
Field::new("c2", DataType::Utf8, true),
Field::new("c3", DataType::Utf8, true),
]);
let batch = RecordBatch::try_new(Arc::new(schema), vec![c1, c2, c3]).unwrap();

// use same predicate for all benchmarks
let predicate = Arc::new(BinaryExpr::new(
make_col("c1", 0),
Operator::LtEq,
make_lit_i32(500),
));
let mut columns: Vec<ArrayRef> = vec![c1, c2, c3];
for _ in 3..column_count {
columns.push(Arc::new(Int32Array::from_value(0, row_count)));

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

do you think it would be useful to use different values for the columns? Maybe since it is a benchmark (not correctness test) all zeros will be fine

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

The values themselves don't really matter for the benchmark. Maybe it's safer to not use identical values just to be sure that the array never gets REE encoded.

}

// CASE WHEN c1 <= 500 THEN 1 ELSE 0 END
c.bench_function("case_when: scalar or scalar", |b| {
let expr = Arc::new(
CaseExpr::try_new(
None,
vec![(predicate.clone(), make_lit_i32(1))],
Some(make_lit_i32(0)),
let fields = columns
.iter()
.enumerate()
.map(|(i, c)| {
Field::new(
format!("c{}", i + 1),
c.data_type().clone(),
c.is_nullable(),
)
.unwrap(),
);
b.iter(|| black_box(expr.evaluate(black_box(&batch)).unwrap()))
});
})
.collect::<Vec<_>>();

// CASE WHEN c1 <= 500 THEN c2 [ELSE NULL] END
c.bench_function("case_when: column or null", |b| {
let expr = Arc::new(
CaseExpr::try_new(None, vec![(predicate.clone(), make_col("c2", 1))], None)
let schema = Arc::new(Schema::new(fields));
RecordBatch::try_new(Arc::clone(&schema), columns).unwrap()
}

fn criterion_benchmark(c: &mut Criterion) {
run_benchmarks(c, &make_batch(8192, 3));
run_benchmarks(c, &make_batch(8192, 50));
run_benchmarks(c, &make_batch(8192, 100));
}

fn run_benchmarks(c: &mut Criterion, batch: &RecordBatch) {
let c1 = col("c1", &batch.schema()).unwrap();
let c2 = col("c2", &batch.schema()).unwrap();
let c3 = col("c3", &batch.schema()).unwrap();

// No expression, when/then/else, literal values
c.bench_function(
format!(
"case_when {}x{}: CASE WHEN c1 <= 500 THEN 1 ELSE 0 END",
batch.num_rows(),
batch.num_columns()
)
.as_str(),
|b| {
let expr = Arc::new(
case(
None,
vec![(make_x_cmp_y(&c1, Operator::LtEq, 500), lit(1))],
Some(lit(0)),
)
.unwrap(),
);
b.iter(|| black_box(expr.evaluate(black_box(&batch)).unwrap()))
});
);
b.iter(|| black_box(expr.evaluate(black_box(batch)).unwrap()))
},
);

// No expression, when/then/else, column reference values
c.bench_function(

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

These are nice labels:

Benchmarking case_when 8192x3: CASE WHEN c1 == 0 THEN 0 WHEN c1 == 1 THEN 1 ... WHEN c1 == n THEN n ELSE n + 1 EN...: Warming up for 3.0000 s

format!(
"case_when {}x{}: CASE WHEN c1 <= 500 THEN c2 ELSE c3 END",
batch.num_rows(),
batch.num_columns()
)
.as_str(),
|b| {
let expr = Arc::new(
case(
None,
vec![(make_x_cmp_y(&c1, Operator::LtEq, 500), Arc::clone(&c2))],
Some(Arc::clone(&c3)),
)
.unwrap(),
);
b.iter(|| black_box(expr.evaluate(black_box(batch)).unwrap()))
},
);

// CASE WHEN c1 <= 500 THEN c2 ELSE c3 END
c.bench_function("case_when: expr or expr", |b| {
// No expression, when/then, implicit else
c.bench_function(
format!(
"case_when {}x{}: CASE WHEN c1 <= 500 THEN c2 [ELSE NULL] END",
batch.num_rows(),
batch.num_columns()
)
.as_str(),
|b| {
let expr = Arc::new(
case(
None,
vec![(make_x_cmp_y(&c1, Operator::LtEq, 500), Arc::clone(&c2))],
None,
)
.unwrap(),
);
b.iter(|| black_box(expr.evaluate(black_box(batch)).unwrap()))
},
);

// With expression, two when/then branches
c.bench_function(
format!(
"case_when {}x{}: CASE c1 WHEN 1 THEN c2 WHEN 2 THEN c3 END",
batch.num_rows(),
batch.num_columns()
)
.as_str(),
|b| {
let expr = Arc::new(
case(
Some(Arc::clone(&c1)),
vec![(lit(1), Arc::clone(&c2)), (lit(2), Arc::clone(&c3))],
None,
)
.unwrap(),
);
b.iter(|| black_box(expr.evaluate(black_box(batch)).unwrap()))
},
);

// Many when/then branches where all are effectively reachable
c.bench_function(format!("case_when {}x{}: CASE WHEN c1 == 0 THEN 0 WHEN c1 == 1 THEN 1 ... WHEN c1 == n THEN n ELSE n + 1 END", batch.num_rows(), batch.num_columns()).as_str(), |b| {
let when_thens = (0..batch.num_rows() as i32).map(|i| (make_x_cmp_y(&c1, Operator::Eq, i), lit(i))).collect();

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

that is a lot of when_thens!

@pepijnve pepijnve Oct 17, 2025

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

Intentionally so. This is a torture test benchmark to really stress the code.

The first 'all reachable' one is really a worst case scenario test case. This is intended to be able to measure improvements in the processing that's being done in each loop iteration. Filtering, scattering, etc.

The second 'few reachable' one is intended to measure the short circuiting behaviour.

let expr = Arc::new(
CaseExpr::try_new(
case(
None,
vec![(predicate.clone(), make_col("c2", 1))],
Some(make_col("c3", 2)),
when_thens,
Some(lit(batch.num_rows() as i32))
)
.unwrap(),
.unwrap(),
);
b.iter(|| black_box(expr.evaluate(black_box(&batch)).unwrap()))
b.iter(|| black_box(expr.evaluate(black_box(batch)).unwrap()))
});

// CASE c1 WHEN 1 THEN c2 WHEN 2 THEN c3 END
c.bench_function("case_when: CASE expr", |b| {
// Many when/then branches where all but the first few are effectively unreachable
c.bench_function(format!("case_when {}x{}: CASE WHEN c1 < 0 THEN 0 WHEN c1 < 1000 THEN 1 ... WHEN c1 < n * 1000 THEN n ELSE n + 1 END", batch.num_rows(), batch.num_columns()).as_str(), |b| {
let when_thens = (0..batch.num_rows() as i32).map(|i| (make_x_cmp_y(&c1, Operator::Eq, i * 1000), lit(i))).collect();
let expr = Arc::new(
CaseExpr::try_new(
Some(make_col("c1", 0)),
vec![
(make_lit_i32(1), make_col("c2", 1)),
(make_lit_i32(2), make_col("c3", 2)),
],
case(
None,
when_thens,
Some(lit(batch.num_rows() as i32))
)
.unwrap(),
.unwrap(),
);
b.iter(|| black_box(expr.evaluate(black_box(&batch)).unwrap()))
b.iter(|| black_box(expr.evaluate(black_box(batch)).unwrap()))
});
}

Expand Down