[sahvx655-wq]

Formatting a long if/elif chain scales quadratically. Profiling a chain of a few thousand elif clauses put nearly all the time in blib2to3's Base.remove, reached from normalize_invisible_parens: after each if/elif/while/for/return keyword it wraps the following test expression in invisible parentheses via wrap_in_parentheses(node, child), which locates the child with child.remove(). The whole chain is a single if_stmt node, so its child list grows with the number of clauses and every remove() rescans it from the start, so wrapping n clauses is O(n^2). This runs in the default stable style and before any line splitting, so it is reachable pre-auth through blackd.

The enclosing enumerate loop already holds each child's position, so I pass it as the index that wrap_in_parentheses gained in #5184 and let it swap the atom in place with set_child instead of remove plus re-insert. Keeping the change in the wrapper call means every clause type flowing through this loop benefits without its own bookkeeping. Output is byte-identical across the test suite and the black plus stdlib sources in stable, preview and skip-string-normalisation modes; a 4000-clause chain drops from ~3.1s to ~1.0s and the curve flattens to linear.

[github-actions]

diff-shades results comparing this PR (b688a84) to main (3224f36):

--preview style: no changes

--stable style: no changes

---

What is this? | Workflow run | diff-shades documentation