GeezerVison

PrettyPrinter._format() now asks _safe_repr() for a container's one-line repr with the remaining line width as a limit, and the list/tuple, dict and dict-view loops stop with a private _TooWide as soon as the repr gets wider than that, so the "does it fit?" check costs at most a line's worth of work per nesting level instead of the whole subtree. The limit is passed through an instance attribute so format()'s signature is unchanged, and it is not applied when format() is overridden, since an override may shorten what it gets from PrettyPrinter.format(); such subclasses keep the current speed. pformat() output, isreadable() and isrecursive() were compared against main on 18,000 random nested structures with various width/compact/expand/depth/sort_dicts options and are identical. A 300-deep list of 20,000 ints goes from 2.7 s to 0.11 s and ordinary wrapped output gets about 20% faster; small objects that fit on one line are a few percent slower from the extra bookkeeping.

Fixes #161114

ANT-2026-1RHKXQEC

• Issue: Speed up pprint on deeply nested containers #161114