Re: pg_dump in 7.4

Philip Warner  writes:
> At 12:39 AM 14/11/2002 -0500, Neil Conway wrote:
>> 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?

> Won't work for functions that build dynamic queries.

That's irrelevant for SQL functions, though, and pg_dump does not need
to worry about dependencies inside other types of functions (since only
SQL functions' bodies are examined during CREATE FUNCTION).

Neil's solution is exactly what I was thinking we should do.  Or almost
exactly, anyway: we don't want to run the rewriter before extracting
dependencies.
		regards, tom lane