Built-in transforms

Everything here is a thin wrapper over the one-row-in, rows-out contract.

Transform	Signature	Purpose
`select`	`select(*cols)`	Keep only the named columns, in the given order.
`drop`	`drop(*cols)`	Remove the named columns.
`rename`	`rename(**old_to_new)`	Rename columns.
`where`	`where(pred)`	Keep rows for which `pred(row)` is truthy.
`derive`	`derive(**col_to_fn)`	Add or replace columns from functions of the row.
`fill`	`fill(**col_to_default)`	Replace empty values with a default.
`dedupe`	`dedupe(*keys)`	Drop repeat rows sharing the same key columns.
`chunk`	`chunk(n)`	Group the stream into lists of up to `n` rows.

Writing your own

If none of the above fit, write a function. Anything that yields rows composes with the built-ins:

from larkspur import Frame

def split_tags(row):
    for tag in row["tags"].split(";"):
        yield {**row, "tag": tag.strip()}

Frame.read_csv("posts.csv") | split_tags | select("id", "tag")