Investigate build and cicd issues

henrifroese commented 3 years ago

There are problems with dependencies (spacy and gensim especially) leading to failing builds.

jbesomi commented 3 years ago

I see, we should probably define the variable name for each library to avoid these kinds of issues also in the future.

henrifroese commented 3 years ago

@jbesomi build issues are now fixed with this pull request, can be merged.

set spaCy to <3.0.0
set gensim>=3.6.0,<4.0
change class HeroTypes(pd.Series, pd.DataFrame) to class HeroTypes without base classes. The base classes are not needed and give a method resolution order error for newer python version
test for both possibilities of equality for categorical results: e.g. when we test kmeans and expect result [1,0,1,0] and kmeans gives us the result [0,1,0,1] (which as 1 and 0 are just categories is functionally the same), we also accept that result

jbesomi commented 3 years ago

Very good job! 🎉 These parts were tricky yet very important! Thanks a lot, Henri 👍 :)

jbesomi / texthero