pepijnve commented on code in PR #19994:
URL: https://github.com/apache/datafusion/pull/19994#discussion_r2745862535
##########
datafusion/physical-expr/benches/case_when.rs:
##########
@@ -517,5 +519,106 @@ fn benchmark_lookup_table_case_when(c: &mut Criterion,
batch_size: usize) {
}
}
+fn benchmark_divide_by_zero_protection(c: &mut Criterion, batch_size: usize) {
+ let mut group = c.benchmark_group("divide_by_zero_protection");
+
+ for zero_percentage in [0.0, 0.1, 0.5, 0.9] {
+ let rng = &mut seedable_rng();
+
+ let numerator: Int32Array =
+ (0..batch_size).map(|_| Some(rng.random::<i32>())).collect();
+
+ let divisor_values: Vec<Option<i32>> = (0..batch_size)
+ .map(|_| {
+ let roll: f32 = rng.random();
+ if roll < zero_percentage {
+ Some(0)
+ } else {
+ let mut val = rng.random::<i32>();
+ while val == 0 {
+ val = rng.random::<i32>();
+ }
+ Some(val)
+ }
+ })
+ .collect();
+
+ let divisor: Int32Array = divisor_values.iter().cloned().collect();
+ let divisor_copy: Int32Array =
divisor_values.iter().cloned().collect();
+
+ let schema = Arc::new(Schema::new(vec![
+ Field::new("numerator", numerator.data_type().clone(), true),
+ Field::new("divisor", divisor.data_type().clone(), true),
+ Field::new("divisor_copy", divisor_copy.data_type().clone(), true),
+ ]));
+
+ let batch = RecordBatch::try_new(
+ Arc::clone(&schema),
+ vec![
+ Arc::new(numerator),
+ Arc::new(divisor),
+ Arc::new(divisor_copy),
+ ],
+ )
+ .unwrap();
+
+ let numerator_col = col("numerator", &batch.schema()).unwrap();
+ let divisor_col = col("divisor", &batch.schema()).unwrap();
+ let divisor_copy_col = col("divisor_copy", &batch.schema()).unwrap();
+
+ group.bench_function(
+ format!(
+ "{} rows, {}% zeros: DivideByZeroProtection",
+ batch_size,
+ (zero_percentage * 100.0) as i32
+ ),
+ |b| {
+ let when = Arc::new(BinaryExpr::new(
+ Arc::clone(&divisor_col),
+ Operator::Gt,
+ lit(0i32),
+ ));
+ let then = Arc::new(BinaryExpr::new(
+ Arc::clone(&numerator_col),
+ Operator::Divide,
+ Arc::clone(&divisor_col),
+ ));
+ let else_null: Arc<dyn PhysicalExpr> =
lit(ScalarValue::Int32(None));
+ let expr =
+ Arc::new(case(None, vec![(when, then)],
Some(else_null)).unwrap());
+
+ b.iter(|| black_box(expr.evaluate(black_box(&batch)).unwrap()))
+ },
+ );
+
+ group.bench_function(
Review Comment:
I'm trying to spot the difference between this benchmark and the one just
above it, they look identical to me. Yet the measurements seem to produce
similar values. What causes that?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
