`sum_rows` calculates sum of columns, `sum_cols` calculates sum of rows about rulinalg HOT 4 CLOSED

Thanks for opening this issue. I'm not sure I can give a good answer to this one - the current implementation is definitely intentional.

I'll try to explain the reasoning:

In the current implementation sum_rows should be read as add all of the rows together. So in your example the matrix has 2 rows, each containing 3 elements. Adding these together should result in a single row with 3 elements. I believe your implementation would make more sense if read as sum along the rows.

I know that this disagrees with numpy for example - though they do explicitly state in the documentation that the sum is along the axis. I think at the very least this confusion should be eliminated in the rulinalg documentation. Or perhaps we should just adopt the along the axis convention.

I'll leave this issue open as I decide how best to handle this...

from rulinalg.

Actually it seems I wasn't remembering numpy's implementation correctly:

>>> import numpy as np
>>> a = np.array([[1,2,3],[4,5,6]])
>>> a.shape
(2,3)
>>> a.sum(0) # sum the rows
array([5, 7, 9])
>>> a.sum(1) # sum the cols
array([ 6, 15])

It seems that we do agree currently?

from rulinalg.

Yes, currently the results match those of numpy.
I didn't look at the numpy implementation at first so I didn't notice that the is sum calculated over the axis instead of along the axis.
Thank you for clarifying!

from rulinalg.

Of course!

In that case I'll close this issue and add a new one for improving the documentation a little. Thanks again for bringing this to my attention.

from rulinalg.