Joiner family parameter names

[Vincent-Maladiere]

Parameters names

During the skrub meeting on 25/09, we discussed bringing more consistency to the parameter names of Joiner-like transformers. The current conclusions are to match pandas API pd.merge with fuzzy_join function to ease adoption while differentiating from pandas for the transformer classes.

This table recaps the current consensus:

name	fuzzy_join	Joiner	InterpolateJoiner	AggJoiner
auxiliary dataframe parameter	left, right	aux_table	aux_table	aux_table
keys parameter	left_on, right_on	main_key, aux_key	main_key, aux_key	main_key, aux_key
suffix parameter	suffixes	suffix	suffix	suffix
match_score parameter	match_score	match_score	NA	NA

Multiple tables classes

We also agreed on only using a single auxiliary table in Joiner, InterpolateJoiner, and AggJoiner. The reason is to make these API easier to use. For each object, we will create a Multi version to deal with multiple tables, respectively MultiJoiner, MultiInterpolateJoiner, MultiAggJoiner. These objects will have the same parameters as their single table version.

Classes names

Also, we voted on naming classes either verbs (e.g. Join) or nouns (e.g. Joiner) throughout all skrub API. The vote leaned on verbs rather than nouns (5 against 4). We can use this poll to continue this discussion and change our votes if we want to.

cc @jovan-stojanovic, @jeromedockes, @myungkim930, @rcap107, @GaelVaroquaux, @flefebv, @LeoGrin, @tigerchen52

[jeromedockes]

as mentioned in 742 we can add the key parameter which can be use in place of main_key and aux_key when they are the same, similar to on in pandas and polars or using in sql