In 9.4.1, I do this:
CREATE TYPE my_test_type as (part1 text, part2 text);
\pset null NULL
WITH test_table(test_col) AS (
VALUES (NULL::my_test_type), (ROW(NULL, NULL)::my_test_type)
)
SELECT *, (test_col).part1, (test_col).part2, test_col IS NULL AS is_null FROM test_table;
And I get this result:
┌──────────┬───────┬───────┬─────────┐
│ test_col │ part1 │ part2 │ is_null │
├──────────┼───────┼───────┼─────────┤
│ NULL │ NULL │ NULL │ t │
│ (,) │ NULL │ NULL │ t │
└──────────┴───────┴───────┴─────────┘
But I expect this result:
┌──────────┬───────┬───────┬─────────┐
│ test_col │ part1 │ part2 │ is_null │
├──────────┼───────┼───────┼─────────┤
│ NULL │ NULL │ NULL │ t │
│ NULL │ NULL │ NULL │ t │
└──────────┴───────┴───────┴─────────┘
Is this expected behavior? I do find references in the docs to input/output of NULL values as components of anonymous
recordtypes, but it's still not clear to me if this would be expected behavior after a cast to a custom type.
Is there a trick to get the result I'm expecting? So far all of the syntactical gymnastics I can think of still produce
thesame result.
I'm a long time PostgreSQL user, but custom types is fairly new to me, so any insight or pointers to appropriate
readingwould be helpful.