ponder-lab / Hybridize-Functions-Refactoring / #2405

81%

main: 83%

Build Type

Pull #750

github

Committed by khatchad

Commit Message

Cover the provably-dynamic decline branch of the numpy shape precondition

Add testNumpyOnDynamicShapeDim: numpy over a reshape(⊤, [-1, 5]) tensor's
dynamic leading dimension. This exercises numpyOverShapeStaticness's DYNAMIC
verdict and the sink's decline path - the soundness-critical branch that no
prior fixture reached (the corpus is all ⊤, the other fixtures static or ⊤).
It also concretely demonstrates the precondition declining a resolved-dynamic
shape-numpy that the blanket-permissive predecessor allowed.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01He1ULx8nRoHzGicMryry1A

Pull Request Pull Request #750: Make the parameter-flow numpy precondition per-dimension shape-aware

Coverage Stats

183 of 202 new or added lines in 3 files covered. (90.59%)

1 existing line in 1 file now uncovered.

2073 of 2552 relevant lines covered (81.23%)

0.81 hits per line