A review article about Numpy was published in Nature. Even though Julia already has superior support for array programming, I think Julia developers can learn something useful from the article.
https://www.nature.com/articles/s41586-020-2649-2
3 Likes
Such as? ( I am genuinely interested in your opinion, this is not trolling.)
1 Like
Such as publishing a review article about one of Julia ecosystem (SciML?) in Science … lol
4 Likes
It’s a good review of Numpy and its history and ecosystem but I guess Julia’s core devs are aware of most of the things in the article. If I were to pick up one thing that Numpy (or Python ecosystem in general) is doing much better than Julia is a good RFC process. Numpy has NumPy Enhancement Proposals (NEP) as discussed in the review and Python has Python Enhancement Proposals (PEP). It’s really nice that they share the discussion and decision process in a format accessible to the people beyond the Python/Numpy community.
As for technical aspects, the article discusses the array protocol for dispatching to different implementations. IIUC, this is a relatively new thing in the Numpy world. But that’s what Julia is way ahead compared to Numpy.
Anyway, I’d rather just celebrate that the long hard work of Numpy developers and contributors for developing and maintaining an OSS is recognized in the scientific world.
18 Likes
on a tangent note, the memes about how “new” the paper is seems to be strong. In this particular tweet, saying Numpy is “jax-like” is a bit like saying “C is C++ -like”
4 Likes
Julia community is pretty ahead (15 years maybe?) in this regard:
(just to list a few I happen to know have citation suggestion)
1 Like
Capturing mindshare may feel like a competition, but it need not be a zero-sum game. Under all of the Twitter flamewars and hacker news rants, there’s a healthy amount of idea cross-pollination across both language communities and the various NumFOCUS projects.
Agreed 100%. There is a lot of tacit knowledge and design discussion that could be reified into something more formal. This is kind of handled in some of the more infamous issues in the Julia repo, but coalescing some of the disparate threads on Slack, GH and Discourse into a central, structured location would reduce some of the redundant back-and-forth.
5 Likes