API Removes the use of fit_ and partial_fit_ in Birch #19297
Conversation
sklearn/cluster/_birch.py
Outdated
| ) | ||
| @property | ||
| def fit_(self): | ||
| return self.__fit |
There was a problem hiding this comment.
Because _fit is already taken.
There was a problem hiding this comment.
A double underscore in Python does "name mangling" which is not necessarily a thing Python developers are familiar with:
https://en.wikipedia.org/wiki/Name_mangling#Python
I don't think this is necessarily a problem for this code but maybe we should avoid them for the sake of not surprising people?
We could use _deprecad_fit_flag instead for instance.
There was a problem hiding this comment.
Updated PR with a non-named mangled name.
jeremiedbb
left a comment
There was a problem hiding this comment.
LGTM.
I guess we still need to document them in the mean time (being done in #18796).
sklearn/cluster/_birch.py
Outdated
| self.__fit, self.__partial_fit = True, False | ||
| return self._fit(X, partial=False) |
There was a problem hiding this comment.
Maybe add a comment "TODO remove in 1.2" so we don't forget (there's another occurence below).
ogrisel
left a comment
There was a problem hiding this comment.
+1 for the general idea. We just need to agree on the point on name mangling raised here: #19297 (comment) and fix it if we judge that it's better to avoid name mangling in our code base.
This PR removes the use of
fit_andpartial_fit_inBirchand update the implementation to not require it.CC @jeremiedbb @adrinjalali