Help Center/ DataArts Fabric/ Developer Guide/ Function Reference/ UDFs/ UDF Development (Python)/ UDTF
Updated on 2025-12-19 GMT+08:00
View PDF
Share

UDTF

The development process for a UDTF is similar to that of a scalar UDF, with the following key considerations:

  • The handler currently only supports returning iterators, so using yield is recommended.
  • UDTFs do not support concurrent execution or vectorized computation.
  • If the handler enters an infinite loop or produces endless output, Fabric SQL will automatically terminate the program based on the timeout setting, which defaults to 1,000 seconds.

Example

  1. Create a UDTF. 
    CREATE FUNCTION py_generate_series(start int, endnum int, step int)
RETURNS TABLE(a int)
LANGUAGE python
RUNTIME_VERSION='3.11.2'
HANDLER='generate_series'
as $$
def generate_series(start, endnum, step):
    current = start
    while current <= endnum:
        yield current
        current += step
$$;
  2. Execute the UDTF. 
    SELECT py_generate_series(1,4,1);
 py_generate_series 
----------------------
                    1
                    2
                    3
                    4
(4 rows)
Parent topic: UDF Development (Python)