fix: schema-qualify DROP INDEX for non-default schemas

[AdhamKhalifa]

Description

Schema-qualify DROP INDEX on PostgreSQL and Oracle so RemoveIndex and partial RemoveConstraint can drop indexes that live outside the connection search_path.

remove_index() already passed a schema-qualified {table} into DROP INDEX "{name}", but the template has no {table} placeholder, so str.format() discarded it. Create still used {table_name} via _qualify_table_name(), which is why the index existed and the drop then failed with index "..." does not exist.

Name-only dialects now qualify the index identifier itself (DROP INDEX "schema"."name"). MySQL and MSSQL keep their table-scoped DROP INDEX ... ON {table} form.

Motivation and Context

Fixes #2288.

How Has This Been Tested?

• make check
• make test (sqlite in-memory): 1959 passed, 148 skipped, 2 xfailed
• Added collect_sql coverage for PostgreSQL, Oracle, SQLite, MySQL, and MSSQL, including a partial unique constraint drop and the no-schema backward-compatible path

Checklist:

• My code follows the code style of this project.
• My change requires a change to the documentation.
• I have updated the documentation accordingly.
• I have added the changelog accordingly.
• I have read the CONTRIBUTING document.
• I have added tests to cover my changes.
• All new and existing tests passed.

Made with Cursor

[codspeed]

Merging this PR will not alter performance

✅ 24 untouched benchmarks

---

_{Comparing AdhamKhalifa:fix/schema-qualify-drop-index (0238f49) with develop (7c2d933)}

[waketzheng]

Thanks for this fix — the root cause analysis is spot-on and the implementation is clean.

The issue was that remove_index() passed a schema-qualified table into DROP_INDEX_TEMPLATE, but name-only dialects (PostgreSQL, Oracle) don't have a {table} placeholder, so str.format() silently discarded it. The new _drop_index_sql() helper handles both template styles by checking for {table} and either keeping the table-scoped form or qualifying the index identifier itself. That's a nice way to avoid per-dialect remove_index overrides, and routing base_postgres.remove_constraint() through the same helper means partial unique constraint drops get the schema qualification for free.

Tests look thorough: schema-qualified drops for PostgreSQL/Oracle, unqualified for SQLite, table-scoped for MySQL/MSSQL, partial unique constraint drop, and the no-schema backward-compatible path. CI is green and CodSpeed shows no impact.

Happy to approve.