[Numpy-discussion] einsum

[Mark Wiebe]

On Wed, Jan 26, 2011 at 6:41 PM, Jonathan Rocher <jrocher at enthought.com>wrote:

> Nice function, and wonderful that it speeds some tasks up.
>
> some feedback: the following notation is a little counter intuitive to me:
>     >>> np.einsum('i...->', a)
>     array([50, 55, 60, 65, 70])
>     >>> np.sum(a, axis=0)
>     array([50, 55, 60, 65, 70])
>  Since there is nothing after the ->, I expected a scalar not a vector. I
> might suggest 'i...->...'
>

Hmm, the dimension that's left is a a broadcast dimension, and the dimension
labeled 'i' did go away.  I suppose disallowing the empty output string and
forcing a '...' is reasonable.  Would disallowing broadcasting by default be
a good approach?  Then,
einsum('ii->i', a) would only except two dimensional inputs, and you would
have to specify einsum('...ii->...i', a) to get the current default behavior
for it.

Just noticed also a typo in the doc:
>
>      order : 'C', 'F', 'A', or 'K'
>         Controls the memory layout of the output. 'C' means it should
>         be Fortran contiguous. 'F' means it should be Fortran contiguous,
> should be changed to
>     order : 'C', 'F', 'A', or 'K'
>         Controls the memory layout of the output. 'C' means it should
>         be C contiguous. 'F' means it should be Fortran contiguous,
>
>
Thanks,
Mark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/numpy-discussion/attachments/20110126/1bcd3e54/attachment.html>