[SciPy-user] [Numpy-discussion] [ANN] NumPy 0.9.6 released
oliphant at ee.byu.edu
Tue Mar 14 12:26:05 CST 2006
Paul Ray wrote:
>On Mar 14, 2006, at 1:33 PM, Travis Oliphant wrote:
>>>Just wondering, does this one also require an update to scipy?
>>>And in general do numpy updates always require an update to scipy,
>>>Or is it only when the numpy C API interface changes?
>>It's only when the C-API interface changes that this is necessary.
>>Pre-1.0 it is likely to happen at each release.
>>After that it is less likely to happen.
>Pre-1.0, I think it is fine for the version numbering to be a free-
>for-all, as it has been. There have been many times, it seems, where
>the latest numpy didn't work with the latest scipy and stuff like that.
The current SVN of numpy and the current SVN of scipy usually work with
each other (perhaps with only a day lag).
>After 1.0, I would like to encourage some discipline with regards to
>numpy version numbering. From 1.0, the API should NOT change at all
This is the plan and it is why we are being slow with getting 1.0 out.
But, we *need* people to try NumPy on their codes so that we can
determine whether or not API additions are needed. Without this testing
we are just guessing based on the developer's own codes which do not
cover the breadth.
>I think that many people with codes that use Numeric or numarray are
>awaiting the numpy 1.0 release as a sign that it is stable and ready
>for prime time, before converting their codes.
We need some of these people to convert earlier than that, though, so we
can make sure that 1.0 really is ready for prime time. I think it's
very close already or the version number wouldn't be so high. We are
just waiting for more people to start using it and report any issues
that they have before going to 1.0 (I'd also like scalar math to be
implemented pre 1.0 as well in case that requires any C-API additions).
Thanks for the feedback. What you described will be the basic behavior
once 1.0 is released.
Note that right now the only changes that usually need to be made are a
re-compile. I know this can still be annoying when you have a deep
More information about the Numpy-discussion