neilconway opened a new pull request, #21366:
URL: https://github.com/apache/datafusion/pull/21366
## Which issue does this PR close?
- Closes #21364.
## Rationale for this change
For `Utf8` and `LargeUtf8` inputs, we can optimize `substr` to avoid copying
the output strings; instead, we can return a `StringViewArray` that points into
the input value buffer.
Benchmarks (M4 Max):
Array args, no count, short strings (strlen=12):
- string_view: 25.8 → 25.2 (-2.5%)
- string: 32.9 → 23.8 (-27.9%)
- large_string: 35.0 → 24.0 (-31.4%)
Array args, with count, long strings (count=64, strlen=128):
- string_view: 41.5 → 41.9 (+1.0%)
- string: 47.4 → 37.0 (-22.0%)
- large_string: 47.6 → 37.5 (-21.2%)
Array args, short count, long strings (count=6, strlen=128):
- string_view: 48.8 → 48.7 (-0.2%)
- string: 59.5 → 49.1 (-17.6%)
- large_string: 65.1 → 49.2 (-24.4%)
Scalar start, no count, short strings (strlen=12):
- string_view: 26.3 → 25.9 (-1.3%)
- string: 32.8 → 24.3 (-25.9%)
Scalar start, no count, long strings (strlen=128):
- string_view: 41.4 → 41.7 (+0.7%)
- string: 47.2 → 37.2 (-21.2%)
Scalar start=1, no count, long strings (strlen=128):
- string_view: 42.0 → 41.9 (-0.2%)
- string: 47.7 → 37.3 (-21.8%)
Scalar args, short strings (count=6, strlen=12):
- string_view: 47.3 → 47.8 (+1.1%)
- string: 61.1 → 50.4 (-17.5%)
Scalar args, long strings (count=64, strlen=128):
- string_view: 42.4 → 42.6 (+0.6%)
- string: 47.7 → 37.8 (-20.8%)
- large_string: 51.5 → 38.4 (-25.5%)
## What changes are included in this PR?
* Implement optimization
* Other minor code cleanup
* Add a benchmark (only somewhat related to this optimization but related to
future optimization work)
## Are these changes tested?
Yes.
## 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]
