Skip to content

feat: N-ary CASE WHEN expressions #20

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

Merged
lukekim merged 3 commits into from
Feb 5, 2026
Merged

feat: N-ary CASE WHEN expressions #20

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
14524f6
N-ary
lukekim Feb 4, 2026
8124f41
fix: update threshold and value literals in n-ary CASE WHEN benchmark
lukekim Feb 5, 2026
7bc7853
refactor: streamline context creation in benchmarks and improve test …
lukekim Feb 5, 2026
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
138 changes: 102 additions & 36 deletions vortex-array/benches/expr/case_when_bench.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,6 @@ use vortex_array::expr::case_when;
use vortex_array::expr::get_item;
use vortex_array::expr::gt;
use vortex_array::expr::lit;
use vortex_array::expr::nested_case_when;
use vortex_array::expr::root;
use vortex_array::session::ArraySession;
use vortex_array::validity::Validity;
Expand Down Expand Up @@ -45,100 +44,167 @@ fn make_struct_array(size: usize) -> ArrayRef {
}

/// Benchmark a simple binary CASE WHEN with varying array sizes.
#[divan::bench(args = [1000, 10000, 100000])]
#[divan::bench(args = [10000, 100000, 1000000])]
Comment thread
lukekim marked this conversation as resolved.
fn case_when_simple(bencher: Bencher, size: usize) {
let array = make_struct_array(size);

// CASE WHEN value > 500 THEN 100 ELSE 0 END
let expr = case_when(
let expr = case_when([
gt(get_item("value", root()), lit(500i32)),
lit(100i32),
lit(0i32),
);
]);

bencher
.with_inputs(|| (&expr, &array))
.bench_refs(|(expr, array)| {
let mut ctx = SESSION.create_execution_ctx();
.with_inputs(|| (&expr, &array, SESSION.create_execution_ctx()))
.bench_refs(|(expr, array, ctx)| {
array
.apply(expr)
.unwrap()
.execute::<Canonical>(&mut ctx)
.execute::<Canonical>(ctx)
.unwrap()
});
}

/// Benchmark nested CASE WHEN with multiple conditions.
/// Benchmark n-ary CASE WHEN with multiple conditions.
#[divan::bench(args = [1000, 10000, 100000])]
fn case_when_nested_3_conditions(bencher: Bencher, size: usize) {
fn case_when_nary_3_conditions(bencher: Bencher, size: usize) {
let array = make_struct_array(size);

// CASE WHEN value > 750 THEN 3 WHEN value > 500 THEN 2 WHEN value > 250 THEN 1 ELSE 0 END
let expr = nested_case_when(
vec![
(gt(get_item("value", root()), lit(750i32)), lit(3i32)),
(gt(get_item("value", root()), lit(500i32)), lit(2i32)),
(gt(get_item("value", root()), lit(250i32)), lit(1i32)),
],
Some(lit(0i32)),
);
let expr = case_when([
gt(get_item("value", root()), lit(750i32)),
lit(3i32),
gt(get_item("value", root()), lit(500i32)),
lit(2i32),
gt(get_item("value", root()), lit(250i32)),
lit(1i32),
lit(0i32),
]);

bencher
.with_inputs(|| (&expr, &array))
.bench_refs(|(expr, array)| {
let mut ctx = SESSION.create_execution_ctx();
.with_inputs(|| (&expr, &array, SESSION.create_execution_ctx()))
.bench_refs(|(expr, array, ctx)| {
array
.apply(expr)
.unwrap()
.execute::<Canonical>(&mut ctx)
.execute::<Canonical>(ctx)
.unwrap()
});
}

/// Benchmark CASE WHEN where all conditions are true (short-circuit path).
#[divan::bench(args = [1000, 10000, 100000])]
#[divan::bench(args = [10000, 100000, 1000000])]
fn case_when_all_true(bencher: Bencher, size: usize) {
let array = make_struct_array(size);

// CASE WHEN value >= 0 THEN 100 ELSE 0 END (always true for our data)
let expr = case_when(
let expr = case_when([
gt(get_item("value", root()), lit(-1i32)),
lit(100i32),
lit(0i32),
);
]);

bencher
.with_inputs(|| (&expr, &array))
.bench_refs(|(expr, array)| {
let mut ctx = SESSION.create_execution_ctx();
.with_inputs(|| (&expr, &array, SESSION.create_execution_ctx()))
.bench_refs(|(expr, array, ctx)| {
array
.apply(expr)
.unwrap()
.execute::<Canonical>(&mut ctx)
.execute::<Canonical>(ctx)
.unwrap()
});
}

/// Benchmark CASE WHEN where all conditions are false (short-circuit path).
#[divan::bench(args = [1000, 10000, 100000])]
#[divan::bench(args = [10000, 100000, 1000000])]
fn case_when_all_false(bencher: Bencher, size: usize) {
let array = make_struct_array(size);

// CASE WHEN value > 1000000 THEN 100 ELSE 0 END (always false for our data)
let expr = case_when(
let expr = case_when([
gt(get_item("value", root()), lit(1_000_000i32)),
lit(100i32),
lit(0i32),
);
]);

bencher
.with_inputs(|| (&expr, &array, SESSION.create_execution_ctx()))
.bench_refs(|(expr, array, ctx)| {
array
.apply(expr)
.unwrap()
.execute::<Canonical>(ctx)
.unwrap()
});
}

/// Benchmark n-ary CASE WHEN with 10 conditions.
#[divan::bench(args = [1000, 10000, 100000])]
fn case_when_nary_10_conditions(bencher: Bencher, size: usize) {
let array = make_struct_array(size);

// Build 10 WHEN/THEN pairs with decreasing thresholds
let expr = case_when([
gt(get_item("value", root()), lit(900i32)),
lit(10i32),
gt(get_item("value", root()), lit(800i32)),
lit(9i32),
gt(get_item("value", root()), lit(700i32)),
lit(8i32),
gt(get_item("value", root()), lit(600i32)),
lit(7i32),
gt(get_item("value", root()), lit(500i32)),
lit(6i32),
gt(get_item("value", root()), lit(400i32)),
lit(5i32),
gt(get_item("value", root()), lit(300i32)),
lit(4i32),
gt(get_item("value", root()), lit(200i32)),
lit(3i32),
gt(get_item("value", root()), lit(100i32)),
lit(2i32),
gt(get_item("value", root()), lit(0i32)),
lit(1i32),
lit(0i32), // else
]);

bencher
.with_inputs(|| (&expr, &array, SESSION.create_execution_ctx()))
.bench_refs(|(expr, array, ctx)| {
array
.apply(expr)
.unwrap()
.execute::<Canonical>(ctx)
.unwrap()
});
}

/// Benchmark n-ary CASE WHEN with 100 conditions.
#[divan::bench(args = [10000, 100000, 1000000])]
fn case_when_nary_100_conditions(bencher: Bencher, size: usize) {
use vortex_array::expr::Expression;

let array = make_struct_array(size);

// Build 100 WHEN/THEN pairs programmatically
let mut children: Vec<Expression> = Vec::with_capacity(201);
for i in (1..=100).rev() {
let threshold = i * 10; // thresholds: 1000, 990, 980, ..., 10
children.push(gt(get_item("value", root()), lit(threshold)));
children.push(lit(i));
}
children.push(lit(0i32)); // else

let expr = case_when(children);

bencher
.with_inputs(|| (&expr, &array))
.bench_refs(|(expr, array)| {
let mut ctx = SESSION.create_execution_ctx();
.with_inputs(|| (&expr, &array, SESSION.create_execution_ctx()))
.bench_refs(|(expr, array, ctx)| {
array
.apply(expr)
.unwrap()
.execute::<Canonical>(&mut ctx)
.execute::<Canonical>(ctx)
.unwrap()
});
}
Loading
Loading