Thread
-
Re: [HACKERS] views and group by (formerly: create view as selec
Tom Lane <tgl@sss.pgh.pa.us> — 1999-05-10T17:31:42Z
Bruce Momjian <maillist@candle.pha.pa.us> writes: > I now get. I am sure it instills confidence in our users: > ERROR: union_planner: query is marked hasAggs, but I don't see any Yes :-(. I've been waiting for Jan to respond to the issue --- I think this is a rewriter problem, so I wanted to know whether he could do anything with it. (See my message "GROUP BY fixes committed" dated 02 May 1999 20:54:30 -0400.) It'd be possible to work around this problem inside the planner, by not believing what the rewriter says about either resnos or hasAggs, but that seems like a kluge rather than a fix. regards, tom lane
-
Re: [HACKERS] views and group by (formerly: create view as selec
Jan Wieck <jwieck@debis.com> — 1999-05-10T18:01:24Z
> > Bruce Momjian <maillist@candle.pha.pa.us> writes: > > I now get. I am sure it instills confidence in our users: > > ERROR: union_planner: query is marked hasAggs, but I don't see any > > Yes :-(. I've been waiting for Jan to respond to the issue --- I think > this is a rewriter problem, so I wanted to know whether he could do > anything with it. (See my message "GROUP BY fixes committed" dated > 02 May 1999 20:54:30 -0400.) > > It'd be possible to work around this problem inside the planner, by not > believing what the rewriter says about either resnos or hasAggs, but > that seems like a kluge rather than a fix. Sorry - forgot about that one. I think the best place to check if the query has aggregates or not is at the beginning of the planner. The rewrite system is recursive, and thus the check at the end of one cycle doesn't guarantee that it will still be true at the end of all rewrites. OTOH views with aggregates are very buggy and introducing ton's of problems (as the other threads show). I'm not sure that it was a good idea to make it partially working :-( for v6.4. There was only the advice that they shouldn't be used. Now it's a released feature. More and more pressure for the subselecting RTE's which are IMHO the only way to solve that all cleanly. The other issue about the resno's is something that I must still search for in the rewriter. There are bad problems where it seems to mangle up not only resno's, it also get's lost of varlevelsup somehow and producing totally wrong varno's into subselects. Still problems coming from the EXCEPT/INTERSECT patch I haven't found so far >:-{ Jan -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #======================================== jwieck@debis.com (Jan Wieck) #