I proposed that we remove sptensor.extract, as the sptensor.__getitem__ functionality should support all of the use cases that we need. Although we proposed to remove sptensor.extract in #160, it is still being used.
Here are the steps that need to be addressed:
[ ] Replace use of sptensor.extract with use of sptensor.__getitem__. Note that the output of extract for sparse tensors with 1 nonzero returns numpy.ndarray, whereas __getitem__ was recently changed to return a scalar in this case.
[ ] Replace use of sptensor.extract in sptensor.__getitem__.
The way that the MATLAB TTB sptensor accessor analog to __getitem__ (i.e., subsref) works is to allow users to include an additional input parameter, 'extract' to extract the result as a vector. I think we should stick with the more numpy-like approach and post process the output of __getitem__ when needed.
I proposed that we remove
sptensor.extract, as thesptensor.__getitem__functionality should support all of the use cases that we need. Although we proposed to removesptensor.extractin #160, it is still being used.Here are the steps that need to be addressed:
sptensor.extractwith use ofsptensor.__getitem__. Note that the output ofextractfor sparse tensors with 1 nonzero returnsnumpy.ndarray, whereas__getitem__was recently changed to return a scalar in this case.sptensor.extractinsptensor.__getitem__.The way that the MATLAB TTB
sptensoraccessor analog to__getitem__(i.e.,subsref) works is to allow users to include an additional input parameter,'extract'to extract the result as a vector. I think we should stick with the morenumpy-like approach and post process the output of__getitem__when needed.