[MATRIX-SIG] reverse of take?

Aaron Watters arw@dante.mh.lucent.com
Wed, 2 Jul 1997 14:37:44 -0400


> > >     a[i, j, k, hint=flavor_woo_woo_17]
> > 
> > I'm reeling about not following this discussion, but I suspect
> > the best approach is for Jim H. to "just do it".  I personally
> > begin to think that excess overloading of indexing is not
> > advisable, possibly confusing.  I vote for "take()" as is with dual
> 
> Until now we haven't discussed much more overloading - just the
> addition of sequence types (including arrays) as indices! And I think
> this kind of operation should be a form of indexing simply because
> people coming from other array languages *expect* it to be.

People keep mentioning other languages to me; I wish I
knew what they were talking about :-).  Seriously, I think
people will whine a bit and then get back to work...

OK, my problem with sequence-generalizes-indices is that
it seems to me that

  Ar[Ar1][Ar2] = x

either won't work properly or will require a change to what
I understand to be the relatively elegant internal representation
for arrays in NumPy as they exist.  I wouldn't recommend
messing up the current implementation for this special case.
The resulting bugs and weird side effects
could be impressive, I think.  If I'm wrong (as I think I may
have been regarding "dual of take" which might be fine in Python)
I take it all back.
-- Aaron Watters

checkout
http://www.chipanalyst.com/report/editorials/edit11_6.html
The number of articles like this one keep piling up...
(but he should have said "geek tragedy" rather than
"Greek tragedy")

_______________
MATRIX-SIG  - SIG on Matrix Math for Python

send messages to: matrix-sig@python.org
administrivia to: matrix-sig-request@python.org
_______________