Re: Adding missing FROM clause - thanks or no thanks?

John Siracusa wrote:
> I recently had a problem where a bunch of postgres backends were taking up
> huge amounts of CPU time. I found a bunch of log messages like this:
>
> NOTICE: adding missing FROM-clause entry for table "b"
>
> Which I eventually tracked to an improperly constructed query that reduced
> to something like this:
>
> select a1, a2 FROM a WHERE a1 = '...' AND b.b1 = '...';
>
> Table b is pretty big and the conditions were such that it was (apparently)
> doing a join of every row in a with every possible row of b. That's
> certainly now what I wanted.
>
> Anyway, I fixed the query construction bug, but I have a question: is this
> "adding missing FROM-clause entry" behavior part of the SQL standard(s) or a
> "feature" of Postgres? Prior to seeing this, I'd have expected a query like
> the above to cause a fatal error in the parsing stage. Put another way, can
> someone give me an example of when this behavior is useful or desirable?

It is a PostgreSQL extension and can be turned off with a
postgresql.conf variable or via SET:

add_missing_from = false

--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 359-1001
+ If your life is a hard drive, | 13 Roberts Road
+ Christ can be your backup. | Newtown Square, Pennsylvania 19073