Alvaro Herrera <alvherre@commandprompt.com> writes:
> FWIW about the new include: I feel a strong dislike about the forward
> declaration you suggest. Defining Relation in elog.h seems completely
> out of place.
Agreed. Maybe a reasonable solution is to allow some ereport helper
functions (or, really, wrappers for the helper functions) to be declared
someplace else than elog.h. They'd likely need to be implemented
someplace else than elog.c, too, so this doesn't seem unreasonable.
The generic helper function approach doesn't seem too unreasonable for
this: elog.h/.c would provide something like
err_generic_string(int fieldid, const char *str)
and then someplace else could provide functions built on this that
insert table/schema/column/constraint/etc names into suitable fields.
regards, tom lane