Re: RFC: listing lock status

Christopher Kings-Lynne wrote:
> Out of interest - why do SRFs need to have a table or view defined that
> matches their return type?  Why can't you just create the type for the
> function and set it up as a dependency?
> 

The only current way to create a composite type (and hence have it for 
the function to reference) is to define a table or view.

We have discussed the need for a stand-alone composite type, but I think 
Tom favors doing that as part of a larger project, namely changing the 
association of pg_attributes to pg_type instead of pg_class (if I 
understand/remember it correctly).

Joe