[Numpy-discussion] field names on numpy arrays

[Fernando Perez]

On Thu, Jun 4, 2009 at 12:28 PM, Pierre GM<pgmdevlist at gmail.com> wrote:
> I foresee serious disturbance in the force...
> When I use structured arrays, each field usually represents a
> different variable, and I may not be keen on having a same operation
> applied to all variables. At least, the current behavior (raise an
> exception) forces me to think twice.

My main use case is really arithmetic: being able to take differences
of structured arrays that contain similar data would make some of our
code here clearer.  But if this doesn't fly, I can always have a
little subtract(a,b) helper that does the 'unpack, subtract,repack'
dance for 'a-b' and similar for any other needed operations.

I realize it's not the most generic case, but we're using structured
arrays a lot for putting data in more manageable/comprehensible
structures, and some basic arithmetic support on the whole array would
be nice sometimes.  The other alternative is to do field-by-field
extractions and operations, which is unnecessary when all the fields
happen to  have the exact same kind of data (numerically speaking, not
conceptually).

> What about the case where you multiply a 1D structured array with a nD
> array ? What should you have ?

I'd punt.  I think it's OK for structured arrays not to support the
full range of binary operations and ufuncs, I was only thinking of
allowing the very simplest: binary ops between absolutely identical
(shape, dtype) arrays whose dtype is an aggregation of native  ones
for which the 'unpack, operate, repack' pattern works.

It seems to me like an improvement in the functionality of structured
arrays, but it's not a big deal.

Cheers,

f