Re: [scikit-image] Numba on pypi

Hi, On Fri, Jul 14, 2017 at 1:29 AM, Ralf Gommers wrote:

On Fri, Jul 14, 2017 at 8:15 PM, Matthew Brett wrote:

...

Hi,

On Thu, Jul 13, 2017 at 9:10 AM, Matthew Brett wrote:

...

Hi,

On Thu, Jul 13, 2017 at 8:50 AM, Ralf Gommers wrote:

...

On Thu, Jul 13, 2017 at 6:16 PM, Stefan van der Walt wrote:

...

Hi everyone,

As many of you know, speed has been a point of contention in scikit-image for a long time. We've made a very deliberate decision to focus on writing high-level, understandable code (via Python and Cython): both to lower the barrier to entry for newcomers, and to lessen the burden on maintainers. But execution time comparisons, vs OpenCV e.g., left much to be desired.

I think we have hit a turning point in the road. Binary wheels for Numba (actually, llvmlite) were recently uploaded to PyPi, making this technology available to users on both pip and conda installations. The importance of this release on pypi should not be dismissed, and I am grateful to the numba team and Continuum for making that decision.

Agreed. Note that there are no Windows wheels up on PyPI (yet, or not coming?). Given that there are no SciPy wheels for Windows either I don't think that that changes your argument much - people should just use a binary distribution on Windows - but I thought I'd point it out anway.

We might be close to a working scipy wheel - discussion evolving over at https://github.com/scipy/scipy/issues/7551#issuecomment-314922271

Following up on my own post - updates on progress for a scipy wheel here:

https://github.com/scipy/scipy/issues/759

...

If we do succeed, that would make the lack of a numba wheel for Windows much more significant.

Does anyone know Continuum's plans in this matter? Is the numba wheel recipe open-source?

Can anyone comment here?

The basic question is - what would happen if Continuum stopped supplying a pypi wheel? If the answer is the standard open source answer - someone else would take over pretty quickly - that's fine. Otherwise, it's a problem.

Can't read their mind, but did look at the build instructions. Doesn't look that hard to build and package, if the need arises (which is unlikely). And the current wheels will not disappear. So I don't really see an issue.

Just FYI - after a lot of hard work over at https://github.com/scipy/scipy/pull/7616 - mostly by Github user Xoviat - we can now build Scipy wheels for Windows. I guess they'll come out at the next Scipy release if not before. Cheers, Matthew