Skip to content

Conversation

@kyleconroy
Copy link
Collaborator

Investigation found that the 2446 explain_todo tests contain
fundamental data inconsistencies that prevent systematic fixes:

  • LIMIT 0/FORMAT stripping varies between similar tests
  • SETTINGS position expectations differ for same syntax
  • AND/OR flattening expectations are inconsistent

The root cause is that explain.txt files were generated from
different ClickHouse versions. Fixes that work for explain_todo
tests break existing passing tests.

Detailed analysis in EXPLAIN_TODO_ANALYSIS.md

Investigation found that the 2446 explain_todo tests contain
fundamental data inconsistencies that prevent systematic fixes:

- LIMIT 0/FORMAT stripping varies between similar tests
- SETTINGS position expectations differ for same syntax
- AND/OR flattening expectations are inconsistent

The root cause is that explain.txt files were generated from
different ClickHouse versions. Fixes that work for explain_todo
tests break existing passing tests.

Detailed analysis in EXPLAIN_TODO_ANALYSIS.md
- Update ClickHouse version to 25.12.1.649 (stable)
- Use ClickHouse server instead of spawning clickhouse local for each query
- Process statements in parallel with configurable worker count
- Add version comment stripping in parser_test.go for compare
- Remove old bash generate_explain.sh script
This test ensures all explain.txt files have the same ClickHouse version
header comment, which guarantees test expectations were generated from
a consistent ClickHouse version.

The test will fail until explain.txt files are regenerated with version
headers using: go run ./cmd/regenerate-explain
@kyleconroy kyleconroy closed this Dec 29, 2025
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.

3 participants