ponder-lab / Hybridize-Functions-Refactoring / #2692

84%

push

github

Surface the depth-limited signal in the evaluator (#840)

* Surface the depth-limited signal in the evaluator

Ariadne 0.52.76 exposes the points-to results its tensor analysis
abandoned because the targeted CFA depth was too short to reach the
relevant context. Collect them per project after tensor analysis,
rendered WALA-free as DepthLimitedPoint (method, value number,
call-string length) so the evaluator needs no visibility into the
embedded WALA packages, and warn when the population is non-empty. The
evaluator emits one row per abandoned result in depth_limited.csv and a
count column in results.csv beside the targeted CFA depth it informs,
so an operator can see which subjects analyzed at an insufficient depth
and raise their per-project targetedCfaDepth accordingly.

Closes #670.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GcqABsog1rtb4m2CDZdcRD

* Clarify that the CAst declaring-class name names the function

Copilot read the declaring-class rendering as a class-not-method
mismatch; in the Python front end each function is modeled as its own
synthetic class, so the class name is the function identity, the same
identification the body-local scans use. Document that instead of
switching to IMethod signatures, whose Python renderings carry
descriptor noise without adding information.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GcqABsog1rtb4m2CDZdcRD

---------

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>

10 of 11 new or added lines in 2 files covered. (90.91%)

3055 of 3651 relevant lines covered (83.68%)

0.84 hits per line