Re: domains, case statements, functions: bug?

On 07/08/2013 06:58 PM, Tom Lane wrote:
> Adrian Klaver <adrian(dot)klaver(at)gmail(dot)com> writes:
>> test=> create function f(t) returns m as $$ select case when true then
>> $1.c end $$ language sql;
>> ERROR: return type mismatch in function declared to return m
>> DETAIL: Actual return type is numeric.
>
> pg_typeof is somewhat helpful here:
>
> regression=# select pg_typeof(t.c) from t;
> pg_typeof
> -----------
> m
> (1 row)
>
> regression=# select pg_typeof(case when true then t.c end) from t;
> pg_typeof
> -----------
> numeric
> (1 row)
>
> The reason for this is that CASE uses select_common_type() to infer the
> output type, and select_common_type intentionally discriminates against
> domain types. The comment therein says:
>
> * If all input types are valid and exactly the same, just pick that type.
> * This is the only way that we will resolve the result as being a domain
> * type; otherwise domains are smashed to their base types for comparison.
>
> So the way to get a CASE to return a domain type is to be sure you
> provide an ELSE with the same result type:
>
> regression=# select pg_typeof(case when true then t.c else null::m end) from t;
> pg_typeof
> -----------
> m
> (1 row)

Found out something new, thanks.

So the following works:

test=> CREATE OR REPLACE FUNCTION utility.f(t)
RETURNS numeric
LANGUAGE sql
AS $function$ select case when true then $1.c else null::m end $function$
;
CREATE FUNCTION

>
> regards, tom lane
>
>

--
Adrian Klaver
adrian(dot)klaver(at)gmail(dot)com