Conversation
Documentation build overview
Show files changed (6 files in total): 📝 6 modified | ➕ 0 added | ➖ 0 deleted
|
|
I need to take some time to use pixi locally before I can review this. I haven't been able to do that yet. |
|
There's also a really good chance that this can close #373. |
|
There are two potentially separate decisions here:
For 1), I don't have strong opinions. Anything that is fast, reliable, maintained, and easy for us to maintain works. If pixi improves on conda in those or other respects I'd like to know how. For 2), we should give more though to who are users are and what they are accustomed to. Installing software with conda is fairly common for computational and biological users. Should that remain the primary/secondary option and pixi the secondary/primary? Supporting two installation pathways leads to a higher maintenance burden, which I don't like. |
|
For 2, we can avoid the 'installing conda/pixi' step entirely hopefully with #332, so I'm not too concerned about it. For the time being, encouraging users to use pixi seems fine. |
|
As for 1, Snakemake itself uses Pixi for CI (and for development!), and unlike Conda, Pixi seems to not have flaky CI. I've had to restart CI in SPRAS for the flaky PCA test, and for flaky conda installations. |
Pixi recently resolved an issue which was blocking my previous attempt at using pixi (one that, while it seemed to work locally, hangs on GitHub CI).
Other than installing
pixi, one just needs to dopixi shellinstead ofconda init ...andconda activate spras. We also do not need to dopip install .anymore, aspixiautomatically installs our localsprasdependency - I'm not sure how well this pairs with #332.I believe the actual infrastructure changes are otherwise minor and fine: the real review comes into making sure we're good on documenting pixi usage.