[adarshsm]

Fixes #885.

_stripws_parenthesis assumed every Parenthesis group is laid out as ( content ), so it indexed tokens[1] and tokens[-2] unconditionally. For "( AS )" the grouping step produces a Parenthesis that holds a single nested Identifier token, so tokens[1] does not exist:

>>> sqlparse.format("( AS )", strip_whitespace=True)
IndexError: list index out of range

>>> stmt = sqlparse.parse("( AS )")[0]
>>> paren = next(t for t in stmt.token_next(0) if isinstance(t, Parenthesis))
>>> len(paren.tokens)
1

Change

sqlparse/filters/others.py:

• When the group is too short to have a bracket on both sides, fall back to _stripws_default instead of indexing past the end.
• Guard the inner trailing-whitespace loop so an inner group whose tokens are all whitespace can't empty itself and then be indexed.

The output for the reported input is now the input unchanged ("( AS )"). I deliberately did not collapse it to "(AS)": the inner content is a single nested group that already went through _stripws_default, and tightening it would need bracket-aware handling that is out of scope for the crash.

Behaviour for well-formed groups is unchanged — SELECT ( a ) FROM t still formats to SELECT (a) FROM t, and the full suite passes.

Tests

Two regression tests in tests/test_regressions.py (near the other issue-numbered tests):

• test_issue885_stripws_malformed_parenthesis — the reported reproducer.
• test_issue885_stripws_short_parenthesis_groups — (), ( ), (a) must not raise.

uv run --group dev --python 3.11 pytest tests/
508 passed, 2 xfailed, 1 xpassed

uv run --group dev ruff check sqlparse/
All checks passed!

A CHANGELOG entry is added under Development Version.