sdf-jkl opened a new pull request, #11924:
URL: https://github.com/apache/arrow-rs/pull/11924

   # Which issue does this PR close?
   
   Relates to #11232. This experimental draft is stacked on its head, 
`4431da4087eeac53f7d15e12def5d2ae8d0a3250`, and includes that PR's changes. 
[Incremental 
comparison](https://github.com/apache/arrow-rs/compare/4431da4087eeac53f7d15e12def5d2ae8d0a3250...sdf-jkl:arrow-rs:codex/variant-path-span-copy-comparison).
   
   # Rationale for this change
   
   Compare an alternative bracket-path escape loop against #11232 on the 
benchmark runner. Local x86-64 measurements were sensitive to instruction 
alignment, so this draft does not claim a portable performance improvement.
   
   # What changes are included in this PR?
   
   Copy each nonempty literal span, append the escaped character immediately, 
and consume adjacent escapes together. Resume span copying at 
`CharIndices::offset()`. Fields without escapes retain #11232's borrowing 
behavior.
   
   The bot comparison will use `4431da4087eeac53f7d15e12def5d2ae8d0a3250` as 
the baseline and this draft's commit as the changed version. The existing 
six-case `variant_path` benchmark is identical on both sides; it does not cover 
the longer strings used in the local investigation.
   
   # Are these changes tested?
   
   Existing path parsing and ownership tests pass. Added cases cover adjacent 
escapes containing multibyte characters and a trailing backslash inside an 
escape run. All `parquet-variant` tests and doctests, crate Clippy, formatting, 
spelling, and the six benchmark smoke cases pass.
   
   A separate differential check matched parsed values and error messages 
against the PR baseline for 3,333,567 inputs. Full-workspace tests and Clippy 
could not run offline because `crc32fast 1.5.3` was not cached; fork CI remains 
to be checked.
   
   # Are there any significant user-facing changes?
   
   No intended semantic or public API changes. Performance remains under 
evaluation.
   
   # AI usage
   
   Codex reconstructed and applied the earlier parser prototype, added the 
focused test cases, ran validation, and prepared this description.
   


-- 
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]

Reply via email to