[Numpy-discussion] Fwd: Advanced selection, duplicate indices, and augmented assignment

[A. M. Archibald]

On 06/01/07, Timothy Hochberg <tim.hochberg at ieee.org> wrote:

> Do we have to revisit that thread? I seem to recall it getting kind of
> cranky. To avoid reliving that, I will attempt dredge up the relevant issue:
>
> "a[indx]+=b" should be the same as "a[indx]=a[indx]+b". All else follow from
> that. If staring at that for a while doesn't enlighten you, then you will
> have to read that thread.

I'm afraid I dredged this issue up previously, and I have no desire to
do so again, but this is not an identity:

In [3]: a = arange(5)

In [4]: a[1:] = a[1:] + a[:-1]

In [5]: a
Out[5]: array([0, 1, 3, 5, 7])

In [6]: a = arange(5)

In [7]: a[1:] += a[:-1]

In [8]: a
Out[8]: array([ 0,  1,  3,  6, 10])

In fact the python documentation simply states that the two should be
"nearly identical". However, a subtler restriction, explained to me by
Robert Kern (http://aspn.activestate.com/ASPN/Mail/Message/scipy-user/3257570),
means that it is in basically infeasible and inappropriate to make
this do what I would have wanted it to.

Nothing, of course, prevents anyone from implementing this in a
function; in fact I think some of the histogram functions are capable
of efficiently doing the most common things one might want this for.

A. M. Archibald

A. M. Archibald