Is there a reason why certain dependencies are pinned in requirements.txt?

[SmokinCaterpillar]

Thanks a lot for dtale, it's a super help and we use it a lot during our development process to investigate data issues with our customers.

I am just wondering if there are any particular reasons why many packages in the requirements.txt are pinned, e.g. scikit-learn==0.24.2. This messes a bit with our workflow, as we use pipenv for dependency resolution. The fixed requirements basically prevent us from correctly listing dtale as a dev dependency because otherwise our docker containers could no longer be built. These fixed dependencies block any dependency resolution with pipenv because it cannot find a valid set of dependencies.

[aschonfeld]

So i had actually been digging into updating scikit-learn to its most recent release but the package “ppscore” is preventing that. If you know of a way i could unpin scikit-learn while also keeping ppscore that would be great.

i was thinking of making ppscore an optional dependency but people seem to use it so i dont want them having to go through hoops to get it

[aschonfeld]

@SmokinCaterpillar ppscore actually has an issue open to upgrade scikit-learn and for some reason they just won't cut a new release even though its been in master for almost 3 months. See here. Please comment that you'd like to see this release cut like I have done and hopefully it will rattle the cage enough 🤞

[aschonfeld]

@SmokinCaterpillar I've integrated the ppscore code directly into the latest D-Tale release (v2.4.0) and thus unpinning sklearn for python environments greater than 3.7. Hope this helps