On Wed, Apr 4, 2012 at 12:02 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
Scott Mead <scottm@openscg.com> writes: > Personally, I feel that if unix will let you be stupid: > $ export PATH=/usr/bin:/this/invalid/crazy/path > $ echo $PATH > /usr/bin:/this/invalid/crazy/path > PG should trust that I'll get where I'm going eventually :)
Well, that's an interesting analogy. Are you arguing that we should always accept any syntactically-valid search_path setting, no matter whether the mentioned schemas exist? It wouldn't be hard to do that.
I think we should always accept a syntactically valid search_path.
The fun stuff comes in when you try to say "I want a warning in these contexts but not those", because (a) the behavior you think you want turns out to be pretty squishy, and (b) it's not always clear from the implementation level what the context is.
ISTM that just issuing a warning whenever you set the search_path (no matter which context) feels valid (and better than the above *nix behavior). I would personally be opposed to seeing it on login however.