ponder-lab / Hybridize-Functions-Refactoring / #3351

88%

push

github

Remember only closure answers that do not depend on the walk's order (#1040)

The side-effect filter asks whether every creation of a written object
lies within the function's closure, and it remembered each method's
answer in `creationsCache`. The walk skips a method it has already
entered, so a method on a call cycle can be answered before its own
walk is complete. Such a cut-short "no" was remembered and then reused
for that method's own question, where it was wrong.

Which method got a cut-short answer depended on the order the
functions were analyzed. That order comes from hashing, so it changes
from run to run even with parallel processing off, and so did the
side-effect verdicts of functions on such cycles.

A remembered answer is now one that does not depend on that order: a
"yes", which found a creating method in the closure, or a "no" from a
walk begun at that method, which saw its whole closure. The caches are
also concurrent now, as `directModCache` already was, though that is
not what caused this.


Claude-Session: https://claude.ai/code/session_01NTnpjPjJj2aZy1ZkUD1eAn

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

5 of 5 new or added lines in 1 file covered. (100.0%)

4795 of 5454 relevant lines covered (87.92%)

0.88 hits per line