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 76829d5ce4 Fix unshredded Variant field extraction benchmark (#11253)
76829d5ce4 is described below
commit 76829d5ce4562616a69ea8b2696e96160514482f
Author: Kosta Tarasov <[email protected]>
AuthorDate: Mon Sep 28 11:11:37 2026 -0400
Fix unshredded Variant field extraction benchmark (#11253)
# Which issue does this PR close?
Follow-up to #10357; corrects the benchmark used by #10358.
# Rationale for this change
The unshredded object fixture has a literal field named `attr.140`, but
`VariantPath::try_from("attr.140")` parses it as two nested fields
(`attr` then `140`). Since the fixture has no `attr` field, the
benchmark returns NULL for every row instead of extracting the intended
value.
# What changes are included in this PR?
Construct the path with `VariantPathElement::field("attr.140")` and
assert that every result is `Some(140)` before entering the timed loop.
# Are these changes tested?
- `cargo bench --offline -p parquet-variant-compute --bench
variant_kernels -- variant_get_unshredded_object_path_262k_rows --test`
- `cargo fmt --all -- --check`
- `git diff --check`
# Are there any user-facing changes?
No. This only corrects the benchmark workload; previous timings for this
benchmark measured missing-path handling.
---
parquet-variant-compute/benches/variant_kernels.rs | 15 +++++++++++----
1 file changed, 11 insertions(+), 4 deletions(-)
diff --git a/parquet-variant-compute/benches/variant_kernels.rs
b/parquet-variant-compute/benches/variant_kernels.rs
index 5a8862f1a9..373a66cdd8 100644
--- a/parquet-variant-compute/benches/variant_kernels.rs
+++ b/parquet-variant-compute/benches/variant_kernels.rs
@@ -15,14 +15,16 @@
// specific language governing permissions and limitations
// under the License.
-use arrow::array::{Array, ArrayRef, BinaryViewArray, BinaryViewBuilder,
StringArray, StructArray};
+use arrow::array::{
+ Array, ArrayRef, BinaryViewArray, BinaryViewBuilder, Int32Array,
StringArray, StructArray,
+};
use arrow::buffer::Buffer;
use arrow_schema::{DataType, Field, FieldRef, Fields};
use chrono::{DateTime, NaiveDate, NaiveDateTime, NaiveTime};
use criterion::{BatchSize, Criterion, criterion_group, criterion_main};
use parquet_variant::{
EMPTY_VARIANT_METADATA_BYTES, Variant, VariantBuilder, VariantBuilderExt,
VariantDecimal8,
- VariantPath,
+ VariantPath, VariantPathElement,
};
use parquet_variant_compute::{
GetOptions, VariantArray, VariantArrayBuilder, json_to_variant,
shred_variant, variant_get,
@@ -198,8 +200,13 @@ pub fn variant_get_unshredded_object_path_bench(c: &mut
Criterion) {
let variant_array =
create_unshredded_object_variant_array(VARIANT_GET_UNSHREDDED_OBJECT_ROWS);
let input = ArrayRef::from(variant_array);
let field: FieldRef = Arc::new(Field::new("typed_value", DataType::Int32,
true));
- let options =
GetOptions::new_with_path(VariantPath::try_from("attr.140").unwrap())
- .with_as_type(Some(field));
+ // The dot is part of the field name, not a separator between nested
fields.
+ let path = VariantPath::from(vec![VariantPathElement::field("attr.140")]);
+ let options = GetOptions::new_with_path(path).with_as_type(Some(field));
+
+ let result = variant_get(&input, options.clone()).unwrap();
+ let result = result.as_any().downcast_ref::<Int32Array>().unwrap();
+ assert!(result.iter().all(|value| value == Some(140)));
c.bench_function("variant_get_unshredded_object_path_262k_rows", |b| {
b.iter(|| variant_get(&input, options.clone()).unwrap())
