etna
etna copied to clipboard
Fix data flow during `fit_transform`
🚀 Feature Request
The current version of data flow during fit_transform is not optimal, it might be improved.
Proposal
- In
to_pandas
method:
- Add flag
copy:bool=True
, implement the corresponding logic - Remove segments from slices
- In
BaseTransform
class:
- Method
fit
should request the reference to the full df - Methods
transform
/inverse_transform
should request the copy of the df with required features
Test cases
- Test that flag
copy
works, get the dataframe, change it and compare with dataframe in dataset - Fix tests for
BaseTransform
if nessesary
Additional context
No response