dimension of cellarrays

[andrescodas]

Hi,

There are issues when converting a cell array with matrices of the same size from matlab to python. The shape of the cell elements are interpreted together with the shape of the cell itself. Then there is a bug on the reshape command.

You implemented a fix for this case for the conversion of structs but not for the cell arrays.

I recommend you to implement the handling of cells in this way:

elif class_name == 'cell':
    out = np.empty(numelems,dtype='O')
    for i in range(numelems):
        cell = libmx.mxGetCell(pm, i)

        if bool(cell):
            out[i] = mxarray_to_ndarray(libmx, cell)
        else:
            ### uninitialized cell
            out[i] = None
      
    out = out.reshape(dims[:ndims], order='F')
    out = out.squeeze()

[mrkrd]

Thanks for opening the issues. I'll check them over the weekend.

[mrkrd]

Hi, There are issues when converting a cell array with matrices of the same size from matlab to python. The shape of the cell elements are interpreted together with the shape of the cell itself. Then there is a bug on the reshape command. You implemented a fix for this case for the conversion of structs but not for the cell arrays. I recommend you to implement the handling of cells in this way: elif class_name == 'cell': out = np.empty(numelems,dtype='O') for i in range(numelems): cell = libmx.mxGetCell(pm, i) if bool(cell): out[i] = mxarray_to_ndarray(libmx, cell) else: ### uninitialized cell out[i] = None out = out.reshape(dims[:ndims], order='F') out = out.squeeze()

Good catch. Could you make a PR with an small test function in test_matlab.py ? Thanks!