Open rth opened 4 years ago
That would be nice, and something I'd like to see happen eventually, but as the recent performance issues showed, there needs to be some performance improvements in sprs
as well.
I've started down this road myself, for a project I'm working on. If anybody else is sufficiently interested, I can put the code into a repo and publish it. I'm happy to write lots of code, but would appreciate guidance and suggestions as to what the API should look like (and of course, coding suggestions -- I'm pretty new to Rust (<3wk)).
And of course, if you want to move such support into sprs, that'd be fine too. I only suggest putting it in a repo as a way to facilitate sharing and discussion.
If anybody else is sufficiently interested, I can put the code into a repo and publish it.
Yes, I think a repo under your account would be a great start. Then later it could be discussed where best to put it.
but would appreciate guidance and suggestions as to what the API should look like
From a Python user perspective, IMO having something that has the scipy.sparse ndarray API even if it doesn't implement all the methods would make it much easier to use in existing code bases. Unless there are reasons specific to sprs to deviate from it. In any case, good support for numpy arrays would be necessary via https://github.com/PyO3/rust-numpy
oops, I forgot you'd opened an issue on this.
python interface for sprs: https://github.com/sparsemat/sprs/issues/315
Hi, have you had a chance to think about how we might make progress on a python wrapper? I threw the thing I wrote onto github, but have somewhat stopped working on it, waiting for feedback.
I am considering to start experimenting with a Python wrapper for sprs, probably in some separate repo. The situation with sparse array libraries is not so great in the Python space right now.
scipy.sparse
is fairly legacy, follows the numpy.matrix protocols instead ofnumpy.ndarray
and generally is difficult to improve because of the amount of code. There is pydata/sparse but it's still fairly new and there are some performance concerns.Just wanted to open the issue to see if there is any interest, or if someone considered it before. It would probably make it easier to compare performance wise with sparse python libraries as well.