On 27.04.21 18:16, Tom Lane wrote:
That's kind of a lot of complication, and inefficiency, for a corner case
that may never arise in practice. We've ignored the risk for default
expressions, and AFAIR have yet to receive any field complaints about it.
So maybe it's okay to do the same for SQL-style function bodies, at least
for now.
Another option would be that we disallow this at creation time.
Don't like that one much. The backend shouldn't be in the business
of rejecting valid commands just because pg_dump might be unable
to cope later.
Since this is listed as an open item, I want to clarify that I'm
currently not planning to work on this, based on this discussion.
Certainly something to look into sometime later, but it's not in my
plans right now.
