Removed version 0.25.0 about pandas-feedstock HOT 8 CLOSED

Won't this give problem for people who pinned to 0.25.0 ?

from pandas-feedstock.

Yep. if someone want the buggy version they need to use the broken label.

from pandas-feedstock.

It seems like removing the 0.25.0 package will cause more problems. All our versions have bugs, and we're not removing all the old ones :)

from pandas-feedstock.

Do you have a case were that is failing? If so people can still access it via the label as I mentioned above. I can restore it to main after the workshop (next week) but I cannot do it before b/c for some reason this version is picked up when people are installing pandas and it breaks all of the tutorials.

from pandas-feedstock.

from pandas-feedstock.

As I mentioned above you need the label:

conda install --channel conda-forge/label/broken pandas=0.25.0

Do you have places were this is causing breakage? If you do need to broken version and the label solution is not good enough can you wait for 2 days? It is a major headache for us now to have all of our tutoriais broken on a cloud instance just b/c of 0.25.0 and asking all the students to upgrade locally to 0.25.1 won't scale well and will eat up the teaching time.

from pandas-feedstock.

Hi @ocefpaf, I just wanted to chime in here to say that moving 0.25.0 to the broken label caused us some problems (malariagen/binder#59). Within our team and in our CI environments we use a pinned conda environment file to ensure reproducibility, both between team members and also if we ever need to go back in time to rerun an old analysis. The conda environment files are tied to the analysis code via github, and so we cannot easily change our installation procedure, and reproducibility is broken if we cannot go back and install an environment that had been used to do some work previously.

I appreciate that in this case there was an urgent need to do something for a workshop, and also in this case there are some bugs in pandas 0.25.0 that means it probably should not be used, but as @TomAugspurger says all versions have bugs, and we had used 0.25.0 successfully to run a bunch of analyses without problems. Can I suggest that there is some tightening up of the conditions for moving a package from main to broken? We are really working hard to achieve reproducibility in our team, and depend heavily on conda and conda-forge to achieve that.

from pandas-feedstock.

It is back already. Re-check your CIs.

from pandas-feedstock.