When get_dummies Shrinks

pd.get_dummies is handy, but it learns nothing. It makes one column for each category in the rows you hand it. A fitted OneHotEncoder always makes the columns it learned during fit.

Task: write column_counts(train_values, batch_values).

• Both lists are values of one category column called "size".
• Count the columns that pd.get_dummies makes from just batch_values (as a one-column DataFrame).
• Fit a OneHotEncoder(handle_unknown="ignore") on train_values, transform batch_values, and count the columns it gives.
• Return [get_dummies column count, encoder column count].