test(sqllogictest): stabilize parquet output_rows_skew with WITH ORDER

[RatulDawar]

Summary

Adds WITH ORDER (x) to CREATE EXTERNAL TABLE skew_parquet / skew_parquet_single in explain_analyze.slt so FileScanConfig preserves scan ordering (preserve_order), keeping per-partition output_rows stable under dynamic file scheduling (PR #21351).

Related

• Follow-up to flaky skew assertions discussed around fix: Skip flaky test in explain_analyze.slt #21866 / Test flake in explain_analyze.slt #21850.

Testing

• cargo test -p datafusion-sqllogictest --test sqllogictests -- explain_analyze (recommended before merge)

Sqllogictest-only change.

Made with Cursor

[2010YOUY01]

Those tests are already stable, and I think the with order statement has no effect.

We have to fix the skipped one. The comments explains the root cause, though I'm not sure about the solution yet.

[RatulDawar]

Yes, I think I worded myself incorrectly. I wanted to improved the coverage as mentioned in the comment by @alamb.
For I checked the code and saw that dynamic file stream task scheduling worked only when preserve_order is false.
So I was thinking that to make the test deterministic we can change tests to trigger preserve order.

[2010YOUY01]

Oh I see, this makes sense. Thank you.

It's not obvious why this fix works, let's add more comments.

[RatulDawar]

@2010YOUY01 made the comment changes

[alamb]

CI failure is a new check added in

• ci: add breaking change detector #21499

Working on resolving it

[rluvaton]

aligned with main since the fix for the detect breaking changes is now resolved, sorry for the trouble

[alamb]

Thanks @rluvaton

[alamb]

Thanks @RatulDawar and @2010YOUY01