ponder-lab / Hybridize-Functions-Refactoring / #1990

80%

Build Type

push

github

Committed by web-flow

Commit Message

Forward a configurable targeted k-CFA depth to the analysis engine (#626)

* Forward a targeted k-CFA depth to the analysis engine (#600 core)

Thread a `targetedCfaDepth` from the refactoring processor to
`EclipsePythonProjectTensorAnalysisEngine` via the 3-arg
`PythonTensorAnalysisEngine(List<File>, String, int)` constructor (Ariadne
0.50.0). The processor field defaults to `MODEL_FORWARD_CFA_DEPTH` (4), the
depth at which the model-forward archetype recovers precise per-context
tensor shapes, with a getter/setter for the evaluator and UI to override.

The full suite passes unchanged at depth 4 (no existing fixture is
depth-sensitive); a depth-sensitive regression test and the evaluator/UI
surfacing follow.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* Surface the targeted k-CFA depth in the evaluator (#600)

Read the depth from the `edu.cuny.hunter.hybridize.eval.targetedCfaDepth`
system property (default `DEFAULT_TARGETED_CFA_DEPTH`), set it on the
processor, and emit it as a `targeted CFA depth` column in the results CSV,
mirroring the other experimental-settings knobs. Hoist the default into a
public `HybridizeFunctionRefactoringProcessor.DEFAULT_TARGETED_CFA_DEPTH`
constant so the processor field and the evaluator share one source.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* Let the test harness forward a targeted k-CFA depth (#600)

Add a `targetedCfaDepth` field (default `DEFAULT_TARGETED_CFA_DEPTH`) and
`setTargetedCfaDepth` to the test harness, applied to the processor it
builds, mirroring how it forwards `inferInputSignatures`. Completes the
configurable-depth surface (engine, processor, evaluator, harness); the
default keeps the suite's behavior unchanged. A depth-sensitive regression
that exercises it will land with the model-forward corpus.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* Pin the targeted-depth soundness gain on a model-forward output (#600)

Add a regression that analyzes the ne... (continued)

Coverage Stats

10 of 10 new or added lines in 2 files covered. (100.0%)

3 existing lines in 1 file now uncovered.

1593 of 1988 relevant lines covered (80.13%)

0.8 hits per line