Model tf.squeeze output shape by dropping singleton axes

[khatchad]

Models tf.squeeze's output shape, the tf.squeeze row of the pass_through audit wala#513 (Bucket 2a) — and the last corpus shape residual for the input-signature evaluation. Stacked on #375 (base branch slice-shape-from-begin-size): crf_forward's state only reaches (2, 4) once #375's tf.slice shape gives (2, 1, 4) to squeeze. Review/merge #375 first; this diff is squeeze-only.

Change

tf.squeeze was a pass_through alias, leaking the input shape unchanged. Replace it with a <new>+<return> summary paired with a dedicated Squeeze generator:

• dtype inherits from input (via PassThroughUnaryTensorGenerator).
• shape drops the singleton axes named by axis (negative axes count from the end), or every statically size-1 axis when axis is absent. Dynamic/symbolic dimensions are never dropped under the axis-absent form (not statically known to be 1); a named non-singleton axis falls back to ⊤ rather than producing a wrong shape.

Tests

• testCrfForward's state (from tf.squeeze(tf.slice(...), [1])) tightens from the pre-squeeze (2, 1, 4) to (2, 4) — the corpus crf_* family is now fully shape-precise.
• New tf2_test_squeeze.py + testSqueezeAxis (tf.squeeze(x, [1]) over (2, 1, 3, 1) → (2, 3, 1)) and testSqueezeAll (tf.squeeze(x) → (2, 3)).
• Full suite green (excluding the unrelated com.ibm.wala.cast.python.jython.test resource bug, fixed in Scope com.ibm.wala.cast.python.jython.test's resource directory #377).

Progresses wala#513.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR improves the TensorFlow modeling in WALA’s Python ML analysis by making tf.squeeze shape-aware instead of treating it as a shape pass-through, enabling tighter inferred tensor shapes in the corpus and in dedicated squeeze tests.

Changes:

• Model tf.squeeze as a dedicated TensorFlow function (<new> + <return>) and route it to a new Squeeze tensor generator.
• Implement shape derivation for tf.squeeze by removing singleton axes (either specified by axis or all statically-known 1 dims when axis is absent), while preserving dtype via PassThroughUnaryTensorGenerator.
• Add/extend tests to pin inferred shapes for both tf.squeeze(x, [axis]) and tf.squeeze(x) and tighten the crf_forward expectation.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
com.ibm.wala.cast.python.test/data/tf2_test_squeeze.py	New TF2 test program exercising tf.squeeze with and without an axis argument.
com.ibm.wala.cast.python.ml/source/com/ibm/wala/cast/python/ml/types/TensorFlowTypes.java	Adds a MethodReference and signature entry for tf.squeeze.
com.ibm.wala.cast.python.ml/source/com/ibm/wala/cast/python/ml/client/TensorGeneratorFactory.java	Routes tf.squeeze calls to the new Squeeze generator.
com.ibm.wala.cast.python.ml/source/com/ibm/wala/cast/python/ml/client/Squeeze.java	New generator implementing squeeze axis resolution and output-shape derivation.
com.ibm.wala.cast.python.ml/data/tensorflow.xml	Replaces squeeze’s prior pass_through alias with a modeled function class and do() method returning a tensor.
com.ibm.wala.cast.python.ml.test/source/com/ibm/wala/cast/python/ml/test/TestTensorflow2Model.java	Updates testCrfForward expected shape and adds testSqueezeAxis/testSqueezeAll.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[codecov]

Codecov Report

❌ Patch coverage is 72.28916% with 23 lines in your changes missing coverage. Please review.
✅ Project coverage is 71.68%. Comparing base (56ccf7b) to head (85d312e).

Files with missing lines	Patch %	Lines
...ce/com/ibm/wala/cast/python/ml/client/Squeeze.java	65.15%	9 Missing and 14 partials ⚠️

Additional details and impacted files

@@                      Coverage Diff                       @@
##             slice-shape-from-begin-size     #379   +/-   ##
==============================================================
  Coverage                          71.68%   71.68%           
- Complexity                          2744     2767   +23     
==============================================================
  Files                                272      273    +1     
  Lines                              20393    20475   +82     
  Branches                            3298     3322   +24     
==============================================================
+ Hits                               14618    14677   +59     
- Misses                              4475     4484    +9     
- Partials                            1300     1314   +14     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.