Re: Does it make sense to break a large query into separate functions?

Merlin Moncure <mmoncure@gmail.com> writes:
> In my experience virtually no useful functions are inlined by the
> planner.   For example, with function:

> create function f(text) returns date as $$
>   select to_date($1, 'YYYY'); $$
> language sql stable;  /* immutable doesn't help */

> I see about 4x time difference between:
> select to_date(v::text, 'YYYY') from generate_series(1,100000) v;
> and
> select f(v::text) from generate_series(1,100000) v;

Something wrong with your measurement technique, because those expand
to identical plan trees for me.

            regards, tom lane