neilconway opened a new pull request, #22283:
URL: https://github.com/apache/datafusion/pull/22283
## Which issue does this PR close?
- Closes #22282.
## Rationale for this change
`split_part` currently uses the Arrow StringBuilder types and computes NULLs
on a per-row basis. This PR switches to using the new bulk-NULL string builders.
Benchmarks (Arm64):
- scalar_utf8_single_char / pos_first: 44.6 µs → 39.1 µs (−11.2%)
- scalar_utf8_single_char / pos_middle: 102.6 µs → 95.8 µs (−6.4%)
- scalar_utf8_single_char / pos_negative: 48.6 µs → 42.5 µs (−12.4%)
- scalar_utf8_multi_char / pos_middle: 134.1 µs → 130.4 µs (−2.9%)
- scalar_utf8_long_strings / pos_middle: 1089 µs → 1101 µs (+1.3%, within
noise)
- scalar_utf8view_long_parts / pos_middle: 140.6 µs → 138.0 µs (−2.0%,
within noise)
- scalar_utf8view_very_long_parts / pos_first: 68.9 µs → 69.4 µs (+1.3%,
within noise)
- array_utf8_single_char / pos_middle: 360.2 µs → 346.6 µs (−3.9%)
- array_utf8_multi_char / pos_middle: 354.3 µs → 343.2 µs (−2.2%,
borderline)
## What changes are included in this PR?
* Switch to new string builder types; compute NULLs in bulk via
`NullBuffer::union_many`
## Are these changes tested?
Yes, covered by existing tests.
## Are there any user-facing changes?
No.
--
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]
