ENH Improving coverage of MinHashEncoder

[jovan-stojanovic]

Fix #371 and part of #376

I replaced the data that needed to be downloaded in the tests with generated ones.

Coverage is up to 98%. Only the _more_tags function is not covered (as in #375)..

Had also to take into account empty strings, which was not the case previously. Linked to #378
Moved up the parameters value verifications to the .fit() method, which seems more logical.

[GaelVaroquaux]

In dirty_cat/_minhash_encoder.py: > - def minhash( Given that this method is not private,

I don't think that it applies, as it's in a private module. Is it exposed in a public module (we renamed all the modules for this specific reason).

[LilianBoulard]

I don't think that it applies, as it's in a private module.

It does apply because minhash is an instance method of the MinHashEncoder, which is part of the public API.

[jovan-stojanovic]

Ok, after some investigation, I think that there is no need to make it private.
Actually, this method is the same as get_fast_hash but using the murmur method. So I renamed it accordingly (as get_murmur_hash) and it can be used as the other methods.. that is applied to any string after fitting the encoder.

[GaelVaroquaux]

This branch has mixed history and the diff is very hard to read.

Maybe you could cherry-pick the relevant commits (potentially in a new branch, if needed)

https://www.atlassian.com/git/tutorials/cherry-pick

[jovan-stojanovic]

This branch has mixed history and the diff is very hard to read.

Maybe you could cherry-pick the relevant commits (potentially in a new branch, if needed)

https://www.atlassian.com/git/tutorials/cherry-pick

Sorry, this was due to the renaming from master to main. It's back to normal now.

[GaelVaroquaux]

@LilianBoulard approved this pull request.

Feel free to merge it, @LilianBoulard. I fear that won't have time to review.

[LilianBoulard]

Sure, LGTM so here we go :)