Make EvalML compatible with the new Woodwork Boolean inference #3892
Conversation
Codecov Report
@@ Coverage Diff @@
## main #3892 +/- ##
=======================================
- Coverage 99.7% 99.7% -0.0%
=======================================
Files 346 346
Lines 36640 36665 +25
=======================================
+ Hits 36510 36532 +22
- Misses 130 133 +3
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here. |
ParthivNaresh
requested review from
eccabay,
jeremyliweishih,
chukarsten,
bchen1116,
christopherbunn and
Cmancuso
There was a problem hiding this comment.
Mostly looks good! Have 1-2 q's but ready to approve once addressed.
I'm curious about is if we need to have a test case for the case you specified in the PR description? E.g. transforming a column of 0 and 1 into boolean values if explicitly set with df.ww.init(logical_types={"col": "Boolean"})
| after_to_before_inference_mapping = { | ||
| new: old for old, new in zip(original_vc.keys(), new_vc.keys()) | ||
| } | ||
| before_to_after_inference_mapping = { | ||
| old: new for new, old in after_to_before_inference_mapping.items() | ||
| } | ||
| if str(y.ww.logical_type) not in ["Boolean", "BooleanNullable"]: | ||
| after_to_before_inference_mapping = { | ||
| old: old for old in after_to_before_inference_mapping.keys() | ||
| } |
There was a problem hiding this comment.
This logic is a bit convoluted but I think I follow it. Can we move lines 159 - 162 up to before 153 and set it as an if/else for the after_to_before_inference_mapping?
There was a problem hiding this comment.
Sure thing, added some comments for clarity too
|
@christopherbunn Added a test to verify |
The Boolean inference has been revised in Woodwork and has been broken into two stages:
Booleanusing this new inference by specifying theBooleanorBooleanNullablelogical types. What does this mean? A column of1s and 0swill not be inferred asBooleanthroughdf.ww.init(), however specifyingdf.ww.init(logical_types={"col": "Boolean"})will transform that column intoTrue and False. This PR is solely to get EvalML compatible with this stage inwoodwork==0.22.0.