Interesting. By this, do you mean select '2020-01-01' :: date; is slow, but select '2020-01-01' ::pg_catalog.date; is fast?
If so, that is very surprising. Any ideas why that is slow?
If so, that is very surprising. Any ideas why that is slow?
It's quite similar to why you fully qualify table names and column names.
Not per se, but it might just as well be that you share the database with several other applications, which might do DDL. Fully qualifying all identifiers is the easiest way to guarantee that you don't have negative dependencies to worry about (i.e. depending on the non-existance of some identifier in a certain schema); the issue will most likely happen only when a) an adversary gets CREATE TYPES access to the database, or when you use custom types and a name that's in use starts to be shadowed.
Examples:
CREATE DOMAIN "bigint" AS pg_catalog.text;
CREATE TYPE "text" AS ENUM ();
Though, in all earnesty, this is also an issue with custom operators:
CREATE OPERATOR = (function = always_false, left_arg = int, right_arg = int);
You can schema-qualify operators ( Col1 OPERATOR(pg_catalog.=) Col2 ), but in doing that you lose operator precedence.
> To be able to create an operator, you must have USAGE privilege on the argument types and the return type, as well as EXECUTE privilege on the underlying function. If a commutator or negator operator is specified, you must own these operators.
(https://www.postgresql.org/docs/14/sql-createoperator.html)