This is an automated email from the ASF dual-hosted git repository.
Jefffrey 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 736090f5e6 Replace BufferBuilder with Vec in decode_binary (#10851)
736090f5e6 is described below
commit 736090f5e6c8d4b03ebe79740e7e5b697ad71917
Author: Xinyao Zhang <[email protected]>
AuthorDate: Wed Aug 26 02:37:02 2026 -0400
Replace BufferBuilder with Vec in decode_binary (#10851)
# Which issue does this PR close?
- Part of #10245.
# Rationale for this change
This replaces one of the remaining `BufferBuilder` callsites listed in
#10245. Using `Vec` for the offsets avoids the builder overhead while
preserving the resulting `ScalarBuffer`.
# What changes are included in this PR?
`decode_binary` now constructs its offsets with `Vec<I>` instead of
`BufferBuilder<I>`.
# Are these changes tested?
Yes:
- `cargo fmt --all -- --check`
- `cargo test -p arrow-row` — 95 unit tests and 8 doc tests passed
- `cargo clippy -p arrow-row --all-targets --all-features -- -D
warnings`
Existing round-trip tests cover the changed path, so no new tests were
added.
Benchmark: `convert_rows 4096 string(100, 0)`
| | Time |
| --- | ---: |
| `main` | 47.769 µs |
| This PR | 45.488 µs |
Criterion reported a 4.4% improvement.
# Are there any user-facing changes?
No.
# AI assistance
Codex was used to identify the callsite, prepare the focused refactor,
run validation, and draft this description.
---
arrow-row/src/variable.rs | 7 +++----
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/arrow-row/src/variable.rs b/arrow-row/src/variable.rs
index 2dfc1807b6..6e19044aef 100644
--- a/arrow-row/src/variable.rs
+++ b/arrow-row/src/variable.rs
@@ -16,7 +16,6 @@
// under the License.
use crate::null_sentinel;
-use arrow_array::builder::BufferBuilder;
use arrow_array::types::ByteArrayType;
use arrow_array::*;
use arrow_buffer::bit_util::ceil;
@@ -284,14 +283,14 @@ pub fn decode_binary<I: OffsetSizeTrait>(
let nulls = decode_nulls_sentinel(rows, options);
let values_capacity = rows.iter().map(|row| decoded_len(row,
options)).sum();
- let mut offsets = BufferBuilder::<I>::new(len + 1);
- offsets.append(I::zero());
+ let mut offsets = Vec::<I>::with_capacity(len + 1);
+ offsets.push(I::zero());
let mut values = MutableBuffer::new(values_capacity);
for row in rows {
let offset = decode_blocks(row, options, |b|
values.extend_from_slice(b));
*row = &row[offset..];
- offsets.append(I::from_usize(values.len()).expect("offset overflow"))
+ offsets.push(I::from_usize(values.len()).expect("offset overflow"))
}
if options.descending {
