With result rows in plpython returned as dicts rather than lists, we ran
into issues with a need to preserve the column order in the resultset.
Of course, dicts in python have an arbitrary, non-random order. It's
consistent in the result but does not match the order in the query. Our
use case was a third party to a customer and ourselves who required a
specific order of columns in CSVs sent to them.
Has there ever been any interest in adding a keyword option for
returning row lists rather than dicts? I couldn't find any, so I
experimented a little and came up with the attached patch. I tested this
a little...managed to make it not segfault at the very least. :) I'm
not even close to a guru so there's probably at least one mistake.
As an example:
test=# do language plpythonu $$
a = plpy.execute("""
SELECT 1 as a, 2 as b, NULL as c, ARRAY[1,2,3] as d;
""", return_list=True)
for row in a:
plpy.notice(repr(row))
$$;
NOTICE: [1, 2, None, [1, 2, 3]]
CONTEXT: PL/Python anonymous code block
DO
I didn't test with Python 3/plpython3u.