[Numpy-discussion] Masked Array for NumPy 1.7

Charles R Harris charlesr.harris at gmail.com
Sun May 20 04:48:55 EDT 2012


On Sat, May 19, 2012 at 11:38 PM, Travis Oliphant <travis at continuum.io>wrote:

>
> On May 20, 2012, at 12:15 AM, Charles R Harris wrote:
>
>
>
> On Sat, May 19, 2012 at 10:48 PM, Travis Oliphant <travis at continuum.io>wrote:
>
>> >
>> > My own plan for the near term would be as follows:
>> >
>> > 1) Put in the experimental option and get the 1.7 release out. This
>> gets us through the next couple of months and keeps things moving.
>> >
>>
>> The "experimental" option does not solve the problem which is that the
>> ndarray object now has masked fields which changes the fundamental nature
>> of an ndarray for a lot of downstream users that really have no idea what
>> has just happened.    I don't see how this has been addressed by any
>> proposal except for the one I have suggested which allows a masked array
>> object and a regular ndarray to co-exist for a time.    I doubt that the
>> proposal actually helps get 1.7 out any faster either as there are multiple
>> experimental APIs that would have to be created to pull it off on both the
>> C and Python level.
>>
>
> So, remove them in 1.8 and try something else. With experimental (say in
> site.cfg), the base array could even be different. I don't see the problem
> here. Think big.
>
>
> I don't think I understand your mental model of this.    Are you saying
> add an experimental flag at the C-level (essentially a #define that
> eliminates any code involving masked arrays unless the define is made at
> compile time?)
>
> It seems like just applying Nathaniel's patch would be a better approach.
>

Do so then. Otherwise I am going to fork.

Chuck
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/numpy-discussion/attachments/20120520/7452a2fd/attachment.html>


More information about the NumPy-Discussion mailing list