Re: [HACKERS] Error in parser with UNIONS.

> I think I may have uncovered an error in the parser.  The
> following is the simplest example that shows the problem.
> Maybe a counter needs to be reset by 'union' or checked
> after select not statement.  I would like to use this syntax
> in my libpq program.  Is this a bug?  Is it already known?

Not already known, and it is a feature for now. I _should_ be able to
get it to work in v6.4, since I have already made changes elsewhere to
do a better job of guessing types in underspecified queries.

For example, in v6.3.2 the following query does not work:

postgres=> select 'a' || 'b' as "Concat";
Concat
------
ab
(1 row)

The underlying reason for the problem is Postgres' conservative approach
to typing and type coersion. I've made changes to make it a bit more
thorough in its matching attempts, and will look at this case soon.

> Would someone please e-mail me the syntax for the
> "explicit cast" the system wants

postgres=> select text 'a' as X
postgres-> union
postgres-> select text 'b';
x
-
a
b
(2 rows)

Note that this is the SQL92-style of specification; you can also use
"'a'::text" rather than "text 'a'". This example was run on something
similar to the current development source tree, but I would expect
v6.3.2 to behave the same way.

                           - Tom