libpysal
libpysal copied to clipboard
review minimum reqs/deps
- review minimum reqs/deps
- xref
- #783
- #782
https://github.com/pysal/mapclassify/issues/225
At the scientific Python dev summit, work will be done around the lower-bound of dependencies, with an eye towards making this consumable into CI so that projects can automate some of this.
Yes please!
I am looking into this now and as a related concern for requirements/support pandana is not yet compatible with Python 3.13^1.
@knaaptime and I discussed this a bit offline recently and it will also be somewhat of a blocker for getting segregation supported with Python 3.13
Since it's not a hard requirement here or in segregation, we will need to add some logic at least to our testing suite to accommodate Python 3.13.
Also, pandana was never added to our plus dependencies here.
It makes me a bit worried that pandana repo looks nearly deserted.
It makes me a bit worried that pandana repo looks nearly deserted.
Agreed. For the PySAL stuff along with other projects.
It makes me a bit worried that pandana repo looks nearly deserted.
Agreed. For the PySAL stuff along with other projects.
Would vendoring pandana into segregation be an option?
I've considered vendoring or maintaining some kind of fork if necessary. The c++ code is stable afaict, the only maintenance is keeping up with the rest of the pydata stack.
but spopt would probably be a better fit
I am not clear on the rules of vendoring, but with the planned sunsetting of spaghetti, perhaps this would be an oppurtune time for a pysal/network redux? With the first addition of course being the pandana stuff?
It would need to be its own package as it has incompatible license with PySAL. So vendoring within any of the existing tools is not an option. It could in principle be just a fork that we maintain but might be worth checking with current maintainers if they're willing to share the keys to the existing repo. That way the minimal maintenance could be done there without a need to fork or vendor the code.
i emailed them once or twice and made a maintenance offer to try and prompt the last release but never heard back. I can reach back out and maybe I'll also drop a line to the package's original author as well
Recently, I have been considering developing a Rust-based (or utilizing any lazy computational resources, or even disk) library that prioritizes performance while offering options to work with SciPy and NetworkX formats for smaller networks. I have never found an open-source library for that purpose. It would be nice to see that under PySAL, and aligns with the pysal/network proposal.
@gegen07 This is absolutely worth a discussion IMHO.
@gegen07 It might be helpful to open a dedicated Discussion for this.