Skip to content

test: dedupe redundant Spark 4.2 TPC-DS plan-stability golden files#4243

Merged
andygrove merged 1 commit intoapache:mainfrom
andygrove:chore/dedupe-spark4_2-goldens
May 6, 2026
Merged

test: dedupe redundant Spark 4.2 TPC-DS plan-stability golden files#4243
andygrove merged 1 commit intoapache:mainfrom
andygrove:chore/dedupe-spark4_2-goldens

Conversation

@andygrove
Copy link
Copy Markdown
Member

Which issue does this PR close?

Closes #.

Rationale for this change

The fallback-chain prune step in CometPlanStabilitySuite was added in #4129 along with a regen for spark3_5, spark4_0, and spark4_1, but the spark4_2 directory (added earlier in #4126) was never re-regenerated against the prune logic. As a result, most query directories under approved-plans-v1_4-spark4_2/ and approved-plans-v2_7-spark4_2/ are byte-identical duplicates of either the spark4_1 directory or the base approved-plans-v1_4 / approved-plans-v2_7 directory. At read time the suite's fallback chain in getDirForTest already resolves to the appropriate copy, so these duplicates only add maintenance cost without changing test behavior.

What changes are included in this PR?

Re-runs ./dev/regenerate-golden-files.sh (all Spark versions) on a clean checkout of apache/main and commits the resulting deletions. Net change: 262 extended.txt files removed across approved-plans-v1_4-spark4_2/ and approved-plans-v2_7-spark4_2/. No other directories changed.

The remaining spark4_2 entries are queries whose plans actually differ from the spark4_1 fallback:

  • approved-plans-v1_4-spark4_2/: q2, q5, q54 (each with .native_datafusion and .native_iceberg_compat variants).
  • approved-plans-v2_7-spark4_2/: 2 entries.

How are these changes tested?

./dev/regenerate-golden-files.sh ran for all five Spark versions (3.4, 3.5, 4.0, 4.1, 4.2) with CometTPCDSV1_4_PlanStabilitySuite and CometTPCDSV2_7_PlanStabilitySuite, all passing. The deletions reflect the suite's pruneDuplicateQueryDirs step, so the same suites continue to pass on the result.

The fallback-chain prune step in CometPlanStabilitySuite was added in apache#4129
along with a regen for spark3_5/spark4_0/spark4_1, but the spark4_2 directory
(added earlier in apache#4126) was never re-regenerated against the prune logic.
Most spark4_2 query directories on main are byte-identical duplicates of
either approved-plans-v1_4-spark4_1 or the base approved-plans-v1_4, and at
read time the suite's fallback chain already resolves to the appropriate
copy.

This commit runs the full regen and prunes the duplicate spark4_2 entries.
Net change: 262 extended.txt files removed across approved-plans-v1_4-spark4_2
and approved-plans-v2_7-spark4_2. The remaining spark4_2 entries (q2, q5,
q54 in v1_4 and 2 in v2_7) are the queries whose plans actually differ from
the spark4_1 fallback.
Copy link
Copy Markdown
Contributor

@mbutrovich mbutrovich left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You love to see it. Thanks @andygrove!

@andygrove andygrove merged commit c799d62 into apache:main May 6, 2026
83 checks passed
@andygrove andygrove deleted the chore/dedupe-spark4_2-goldens branch May 6, 2026 19:38
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants