Transform Reference (v0)

Transform	Required params	Optional params	Example
cast	types mapping {column: type}	on_error: fail (default) | null | keep	Transform("cast", params={"types": {"age": "integer"}, "on_error": "null"})
select	columns: list of column names	—	Transform("select", params={"columns": ["person_id", "age"]})
derive	new: column name, expr: expression string	overwrite (bool, default False), strict (bool, default True)	Transform("derive", params={"new": "is_adult", "expr": "age >= 18", "overwrite": True})
filter	where: expression string	strict (bool, default True)	Transform("filter", params={"where": "age >= 30 and country == 'KE'"})
drop	columns: list of column names	—	Transform("drop", params={"columns": ["debug_col"]})
validate	—	sample_rows (int, default 5000), fail (bool, default True), strict_schema (bool, default True)	Transform("validate", params={"sample_rows": 1000, "fail": False})
join	right: URI or descriptor, on: column/list of columns	how (inner default/left), right_on, suffixes (("_left","_right") default), options (dict)	Transform("join", params={"right": "other.csv", "on": ["id"], "how": "left"})

Notes: - Expression params (expr, where) use the same DSL as filter/derive (logical ops, comparisons, literals, column names). - Types accepted by cast align with frictionless types (integer, number, string, etc.). - validate requires the optional frictionless dependency to be installed.

YAML Ergonomics

• Canonical cast.on_error values are strings: fail, null, keep.
• Unquoted YAML null for on_error is accepted and normalized to "null".
• Unquoted YAML key on: in join params is accepted even when parser treats it as a boolean key.

Expression DSL

Used by filter.where and derive.expr:

• logical: and, or, not
• comparisons: ==, !=, >, >=, <, <=
• literals: strings, numbers, booleans, null
• parentheses for grouping
• arithmetic in derive: +, -, *, /