Changing columns

Each operation returns a new frame; use the returned frame for the next step.

Function	Signature	Behavior	Failure
with_column	(df: Frame[n], name: string, col: Column[n]) -> Frame[n]	Replaces the column name in place, or appends it as the last column	None
mutate	same as with_column	Same as with_column	None
rename	(df: Frame[n], old_name: string, new_name: string) -> Frame[n]	Renames in place; the column keeps its position	rename: target column already exists; missing column: OLD
drop_column	(df: Frame[n], name: string) -> Frame[n]	Removes the column	drop_column: missing column NAME

module Coral.BookMutation
import Coral.Frame (FloatCol, from_pairs, with_column, rename, drop_column, int_col_of_list, ncols)
export (main)
def main() -> i64 = {
  frame = from_pairs([("price", FloatCol(to_tensor([10.0f32, 20.0f32])))])
  renamed = rename(frame, "price", "cost")
  extended = with_column(renamed, "qty", int_col_of_list([2i64, 3i64]))
  trimmed = drop_column(extended, "cost")
  ncols(trimmed)
}

main returns 1: only qty remains.

Pitfalls

• Column length. Give with_column a column with the frame's row count. It does not check the length when it runs, so a column of a different length produces a frame whose columns disagree.
• Replacing changes the type. with_column replaces a column of any type with the new one; replacing a FloatCol a with an IntCol keeps a at its position as an integer column.
• Renaming to the same name. rename(df, "a", "a") fails with rename: target column already exists, because the target check runs first.
• Native builds. chelis build rejects programs that call drop_column; see limitations.