Pandas: hard or soft dependency

[adrinjalali]

Right now in our CI we install install and doc dependencies, and the latter includes pandas.

However, we haven't assumed pandas as a hard dependency and users can use the library w/o having it installed.

#45 for example, assumes it's okay for it not to be installed, if we want to go down this path, we should also test it, or we can make it an install dependency and assume it's always installed.

Which path should we take? We could also assume it a hard dependency now, and only make it soft later.

[BenjaminBossan]

Why is pandas a doc requirement? It doesn't seem to be used. If so, is it not easier to remove the dependency for now? We should probably have it as a tests dependency though, so we can check if data frames work as input.

[adrinjalali]

the docs use it with as_frame=True. If we want to test it, we should have some CI with and some CI w/o to make sure both work.

[BenjaminBossan]

the docs use it with as_frame=True

Ah I see, that's easy to miss.

If we want to test it, we should have some CI with and some CI w/o to make sure both work.

I think we should eventually test it. Probably there is a possibility to patch the import to make it act as if pandas were not installed.

[adrinjalali]

#98 tests this. Closing then. Keeping pandas as a soft dependency.