Re: recent ALTER whatever .. SET SCHEMA refactoring

Kohei KaiGai escribió:

> I'm probably saying same idea. It just adds invocation of external
> functions to check naming conflicts of functions or collation; that
> takes additional 4-lines for special case handling
> in AlterObjectNamespace_internal().

Okay, I can agree with this implementation plan. I didn't like your
patch very much though; I'm not sure that the handling of collations was
sane. And the names of the AlterFooNamespace_oid() functions is now
completely misleading, because they no longer do that. So I renamed
them, and while at it I noticed that the collation case can share code
with the RENAME code path; the function case could probably do likewise,
but it becomes uglier.

Note that after this patch, the error message now cites the function
signature, not just the name. This is more consistent with other cases.

9.2 and HEAD:
alvherre=# alter function foo.f(int,int) set schema bar;
ERROR: function "f" already exists in schema "bar"
(this error is misleading, because there can validly be a f(int)
function in schema bar and that wouldn't cause a problem for this SET
SCHEMA command).

After my patch:
alvherre=# alter function foo.f(int,int) set schema bar;
ERROR: function f(integer, integer) already exists in schema "bar"

--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services