apache / apache/datafusion

CrossJoinExec merges schema metadata inconsistently with the logical plan

Open
#23,434 1 comment 0 reactions 1 assignee Claimed by @gmhelmold View on GitHub
bug
Dominant language
Rust
Stars
9.3k
Forks
2.4k
Avg merge
3d 7h
Merged PRs (30d)
344

Description

### Describe the bug

When two inputs carry the same schema-level metadata key with different values, a pure cross join (CrossJoinExec) over them feeding an aggregate fails with:

Internal error: Physical input schema should be the same as the one converted
from logical input schema. Differences:
- schema metadata differs: (physical) {"pandas": "RIGHT-SCHEMA"} vs (logical) {"pandas": "LEFT-SCHEMA"}

Hash/equijoins and filtered nested-loop joins on the same tables work; only the pure cross-product path is affected.

### To Reproduce

_No response_

### Expected behavior

_No response_

### Additional context

xref: #16221 #21127

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.