haochunchang opened a new pull request, #26202:
URL: https://github.com/apache/datafusion/pull/26202
## Which issue does this PR close?
- Closes #14436.
Continues #20720 by @buraksenn, rebased onto current main.
## Rationale for this change
When a query repeats a CTE name, or a table name or alias in the same `FROM`
scope, the error has no source location. For relations it also comes late, as
`Schema contains duplicate qualified field name`, so users can't easily see
which two names collide.
## What changes are included in this PR?
- Duplicate CTE names: attach a `Diagnostic` at the repeated name, with a
note
at the first definition.
- Duplicate relation names in a `FROM` scope (comma-separated entries, JOINs,
parenthesized joins):
- Each one adds a non-fatal warning `Diagnostic`
(`table name "t" specified more than once`, with a note at the first
definition). There is exactly one warning per duplicate.
- If planning then fails with `DuplicateQualifiedField` for that name, the
error gets a `Diagnostic` pointing at both relations. The error message
text is unchanged.
- Whether a query plans or fails is unchanged.
- Names are compared as `TableReference`s that match the planned qualifiers.
A quoted alias `"s.t"` does not match `s.t`, and an unaliased table
function `foo(..)` (qualifier `foo()`) does not match a table `foo`. Names
the check can't resolve are skipped.
## What is the testing strategy for this PR?
- `datafusion/sql/tests/cases/diagnostic.rs`:
- `test_duplicate_cte_name`: error Diagnostic spans for duplicate CTEs.
- `test_duplicate_table_name`: for each FROM shape, checks that there is
one
warning with the right spans. Queries with overlapping columns fail with
the original message plus an error Diagnostic. Queries with disjoint
columns and LEFT SEMI joins plan successfully.
- `test_duplicate_table_name_single_warning_in_multi_entry_from`: one
warning, not two, when a duplicate spans comma-separated entries.
- `test_same_table_name_in_different_scopes`: no warning across scopes.
- `datafusion/sqllogictest/test_files/join.slt.part`: existing expectations
are
unchanged. New cases cover a quoted dotted alias vs a schema-qualified
table,
and a table function vs a same-named table.
## Are there any user-facing changes?
- New warning `Diagnostic`s for duplicate relation names. Callers read them
with `SqlToRel::take_warnings`.
- An error `Diagnostic` is now attached to duplicate CTE and duplicate
qualified field errors.
- No breaking changes: the same queries plan or fail as before, with the same
error messages.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
--
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]
