Timo Kaufmann

Rebased, now that #118 is merged.

I have continued work on top of #119 to avoid merge conflicts. We're getting very close to passing the `default_constructible` test! Besides https://github.com/scikit-learn/scikit-learn/issues/17756, with the fixes in this PR and...

All our estimators are default constructible now and the test passes with https://github.com/scikit-learn/scikit-learn/pull/17936 :tada: That means that we cannot add the test before the PR is merged and a new...

The PR was merged upstream. I'm not sure if it'll make it into 0.23.3 or if we have to wait for 0.24. Either way, there should be a stable version...

After #154 and #155, this will be the status of the "no attributes in init" check: ``` AssertionError: Estimator should not set any attribute apart from parameters during init. Found...

Good news and bad news. Good news: After #159, we are now compliant with "no attributes in init" :tada: Bad news: Basically all other checks are unusable as long as...

Just to re-state the problem (we talked about this in some other PR/issue, but I can't remember which): According to scikit-learn, `fit` should expect an `X` of the shape `(n_samples,...

I tend to agree. I will look into writing a wrapper anyway, which will at least enable us to make use of the remaining checks.

Okay, I have done that. Just for rankers for now, I'll fix the checks there first (of course they don't all pass). I already added some little fixes to this...

Now that we use poetry & nix, I have included the environment with the patched scikit-learn that is necessary to run these tests. A simple `nix-shell` will load it (though...