Overloaded `plus` method bug about mappedtensor HOT 3 CLOSED

This is an issue with an incorrect call to SliceFunction. The arguments are not specified correctly.

from mappedtensor.

I've been thinking about the plus operator for MappedTensor. It seems there is no advantage of implementing it as an overloaded operator, like there is with multiply (which is a trivial operation). Most operations that require a full load of the tensor should probably be done only as necessary on a portion of the tensor being used at that moment, e.g. a(:,1,2,3)+1.

There seem a couple of interesting paths to me:

1. Remove the + operator entirely and keep the class as clean as possible.
2. Implement an eval string that remembers any operations done on the MappedTensor in the right order and applies them at the point they are requested.

So line 451:

tfData = mtVar.fRealFactor .* mt_read_data(mtVar.hShimFunc, mtVar.hRealContent, S, vnReferencedTensorSize, mtVar.strStorageClass, mtVar.nHeaderBytes, mtVar.bBigEndian, mtVar.hRepSumFunc, mtVar.hChunkLengthFunc);

might become a recursive set of feval() function calls done in the right order, e.g. 'minus( divide( plus( times( power( mtVar, A), B), C), D), E)'.

Option 1) is easier and probably the way to go but I can see the merit in 2) as well. The current approach is kind-of in between :)

from mappedtensor.

Yeah, I'm thinking that the plus method isn't so useful. I agree with your point, that anything requiring a full load of the tensor should be performed on a referenced portion. I'll change it to raise an error.

from mappedtensor.