andygrove opened a new pull request, #4833: URL: https://github.com/apache/datafusion-comet/pull/4833
## Which issue does this PR close? N/A ## Rationale for this change The documentation for window functions was out of date and contradicted itself. The compatibility guide (`compatibility/operators.md`) and the contributor roadmap still stated that native window execution is "disabled by default and known to be incorrect" and "should not be used in production", citing the epic [#2721](https://github.com/apache/datafusion-comet/issues/2721). That epic, along with the other referenced issues (#2705, #2841, #4007), is now closed. In reality, `WindowExec` runs natively and `spark.comet.exec.window.enabled` defaults to `true`. A broad set of window functions is accelerated, and any shape Comet does not handle falls back to Spark rather than producing an incorrect result. The `expressions.md` and `operators.md` pages already reflected this, so the guide and roadmap were the stale pages that disagreed with them. ## What changes are included in this PR? - Rewrite the "Window Functions" section of `compatibility/operators.md` to describe the current behavior: the ranking, value, and aggregate window functions that are accelerated, the supported frame units and bounds, and the specific cases that fall back to Spark (statistical aggregates as window functions, unsupported input types, decimal sliding-frame `sum`/`avg`, `RANGE` frames with date/decimal offsets, `first_value`/`last_value` on `RANGE` with a literal offset, non-literal `lag`/`lead` defaults, `GROUPS` frames, and `DISTINCT` over a window). - Update the `WindowExec` note in `operators.md` to drop the closed-issue reference and point to the window compatibility section. - Tighten the `window_funcs` prose in `expressions.md` to note that statistical aggregates run natively as plain aggregations but fall back as window functions, and to link to the compatibility guide. - Rewrite the "Window Expressions" roadmap section, which cited only closed issues, to describe what is accelerated today and the genuinely remaining gaps. - State the unsupported aggregate input types positively (the Spark types Comet does not accept) rather than by exclusion. ## How are these changes tested? Documentation-only change. Formatting was verified with `prettier --check`. -- 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]
