Bruce Momjian <maillist@candle.pha.pa.us> writes:
> Do you want to be making these commits now? We haven't split the tree,
> so they will appear in 6.5.x.
Yeah, I know. I'm in bugfix mode: CASE does not work in any context
involving GROUP BY or aggregates, egselect coalesce(f1,0) from int4_tbl group by f1;ERROR: Illegal use of aggregates
ornon-group column in target list
Since we now support CASE "officially", I think this is important
enough to fix in 6.5.*. The cause is that parse_agg.c's routines
neglected the CaseExpr nodetype case. I couldn't quite stomach
adding the same boilerplate code to yet another place, so I decided
to start implementing the suggestion I made a while ago to create
centralized tree-recursion logic.
regards, tom lane