This helper function applies fn to each of col_names and assigns the
results back into new_data in bulk. It should be used in bake.step_*()
functions instead of assigning to new_data[[col_name]] inside a for loop,
which is quadratic in the number of columns and becomes prohibitively slow
for wide data.
Arguments
- new_data
A tibble.
- col_names
A character vector, denoting columns to transform.
- fn
A function taking one to three arguments: the column to transform, the position of that column within
col_names, and its name. Only the arguments thatfnaccepts are passed, so\(x) x * 2,\(x, i) x - means[[i]], and\(x, i, col_name) ...are all valid. It must return a vector the same length as the column it is given.
Details
Use the position i to look up per-column values estimated during prep(),
after aligning them to col_names once outside of fn. Do not use
col_name to index a named vector with one element per selected column, as
the repeated name matching is what makes the naive approach slow. col_name
is intended for error messages and for lookups into small objects.