Naming things across the project

[Vincent-Maladiere]

This discussion is not a priority for the release.

The inputs of internal functions and methods currently use a mix of different variable names, which is confusing. The documentation also mixes concepts.

Should we use:

Questions 1: dataframe vs table vs X

• table refers to the SQL world
• dataframe refers to Pandas, Polars, and the dataframe API
• X is less clear; it refers to both numpy arrays and dataframes alike in scikit-learn

Questions 2: column_name vs column (as the name of the column)

• column_name is used in the dataframe API
• column is used in Pandas and Polars

Question 3: column vs series (as the data container)

• column is used in the dataframe API
• series is used in Polars and Pandas

---

On the website:

• Prepping tables for machine learning
• Easy learning on pandas dataframes

Publicly exposed methods/functions are mostly fine, except maybe deduplicate:

to_datetime(X)
DatetimeEncoder().fit(X)
{Joiner, AggJoiner, AggTarget, InterpolationJoiner}()(aux_table)
{Joiner, AggJoiner, AggTarget, InterpolationJoiner}.fit(X)
fuzzy_join(left, right)
deduplicate(data)
GapEncoder().fit(X)
MinHashEncoder().fit(X)
SimilarityEncoder().fit(X)
{Select, Drop}Cols().fit(X)
compute_ngram_distance(unique_words)

Internal functions and methods are where this discussion is helpful.

What would you prefer to see internally in skrub and on the website?