Tom Lane <tgl@sss.pgh.pa.us> writes:
> * We don't store dependencies for SQL functions to things mentioned in
> the SQL function body. (Maybe we should, but we don't.) So there's
> data missing in that case, and possibly other cases.
This might be interesting to do, and we could tie it into the need to
invalidate PL/PgSQL functions that depend on a database object when
the object is changed.
Perhaps when the function is defined, we run all the SQL queries in
the function body through the parser/analyzer/rewriter, and then
generate dependencies on the Query trees we get back?
In any case, there would be a limit to what we could divine from the
function definition (e.g. we'd get practically no info about a
function defined in C) -- but this might make things a little nicer,
anyway.
Cheers,
Neil
--
Neil Conway <neilc@samurai.com> || PGP Key ID: DB3C29FC