[sahvx655-wq]

Description

Profiling --preview formatting of a module with a long run of implicitly concatenated f-strings (x = f"{a0}" f"{a1}" ...) showed update_sibling_maps taking most of the runtime, with Node.replace close behind. The cause is that replace rebuilds the parent's whole children list and then calls invalidate_sibling_maps, so the next read of a sibling rebuilds the prev/next maps in full. visit_fstring/visit_tstring call node.replace(string_leaf) once per concatenated piece, so on a node with many children every replacement costs O(children) for the list rebuild and another O(children) for the map rebuild, and the whole pass becomes quadratic. A 4000-piece concatenation takes about 5.9s and the growth is plainly quadratic; this is reachable from untrusted source and from unauthenticated blackd with X-Preview.

#5178 made the other mutators (remove/set_child/insert_child/append_child) update the sibling maps incrementally, but replace was left invalidating them wholesale, and it is also the only mutator that rebuilt the entire children list. The fix locates the replaced child once, splices the new children in with a list slice, and adds _replace_child_in_sibling_maps to patch the cached maps in place (handling the one-to-many case) with the same fallback-to-invalidate the neighbouring helpers already use. Keeping the splice next to the list edit means a sibling read after a replace no longer pays for the whole node. Output is byte-identical across the test suite (465 passed) and the Black source tree in stable and preview, and a randomised map-equivalence check confirms the spliced maps match a fresh rebuild; the 4000-piece input drops from ~5.9s to ~1.1s and the curve flattens.

Checklist - did you ...

• Implement any code style changes under the --preview style, following the stability policy? (no style change; output is identical)
• Add an entry in CHANGES.md if necessary?
• Add / update tests if necessary? (no behaviour change to assert; exercised by the existing suite, same as avoid rebuilding sibling maps on every node mutation #5178)
• Add new / update outdated documentation?

[github-actions]

diff-shades results comparing this PR (b983801) to main (46c2abc):

--preview style: no changes

--stable style: no changes

---

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