No error retrieval for tensor analysis

[khatchad]

We should call https://github.com/ponder-lab/ML/blob/870a2f509a5acc4a81f34979854d32108eb54323/com.ibm.wala.cast.python.ml/source/com/ibm/wala/cast/python/ml/client/PythonTensorAnalysisEngine.java#L729-L731. Here's an example: https://github.com/ponder-lab/ML/blob/870a2f509a5acc4a81f34979854d32108eb54323/com.ibm.wala.cast.python.ml/source/com/ibm/wala/cast/python/ml/driver/PythonDriver.java#L213

We should call it here:

Hybridize-Functions-Refactoring/edu.cuny.hunter.hybridize.core/src/edu/cuny/hunter/hybridize/core/refactorings/HybridizeFunctionRefactoringProcessor.java

Line 437 in 0317611

TensorTypeAnalysis analysis = engine.performAnalysis(builder);

[khatchad]

Ah, these aren't programming errors. They're errors in the analyzed code.

[khatchad]

Actually, this affects the tensor inference per wala/ML#195. In other words, if this set is non-empty, currently, tensor shape errors may affect the tensor type inference.

[khatchad]

Since we are using a "best-effort" analysis, I don't think we would fail here but maybe add a refactoring warning or info.