On Mon, Oct 26, 2009 at 12:01 AM, Tom Lane <t...@sss.pgh.pa.us> wrote:
> Timothy Madden <terminato...@gmail.com> writes: > > What I want is compatible with existing code and the current default > > behavior. Just look for a LANGUAGE SQL declaration in the function > > header (before the body). > > > If found expect the in-place definition of the function body to follow. > > If not found expect a string literal that holds the function body to > follow, > > with the LANGUAGE declaration after (default behavior). > > This proposal is unfortunately complete nonsense, because it fails to > address the question of how you figure out where the function body *ends*. > We have to have a simple and not-language-specific rule for that. Even > if the backend could be made smart enough to handle a variety of cases, > we could hardly expect client-side code (like psql) to track all the > cases. And psql does need to understand where the CREATE FUNCTION > command ends, so that it can tell when to ship the command off to the > backend. > By the standard the routine body is a <SQL procedure statement> and the question of how to figure out where the function body ends should be answered as such. I am talking about two cases, the one psql already handles, and the one where the body is (and ends as) a <SQL procedure statement>, which statement again psql should already understand and which is signaled by the LANGUAGE SQL declaration in the function header. Thank you, Timothy Madden
