Thanks Krisztian,
I checked google and most of the pgsql archive messages it pointed to, but I
found that this message is commonly related to subselects in table
constraints.
If I didn't misunderstand it, Tom once titled this as "probably solved in
6.5" or such, back in 1999.
However, my problem is different, and even if it's not vital now, I'd like
to get some suggestions where to search if the problem persists.
I use 7.2.1 and met the problem BUT ONCE, after changing a field definition
in a view from a function call to the explicit content of the function, like
this:
-- t and x are business files,
-- with items referring to t in t_item
-- and their N:M relations with x and x_item (not important now) in t_x.
-- I want to select the latest x.id from the relations table:
--
-- myfunc(t.id, t_item.no) AS x, -- this was the original
(SELECT x FROM t_x
WHERE t_x.t = t.id AND t_x.t_item = t_item.no
ORDER BY tstamp DESC LIMIT 1
) AS x,
The error message occured after I recreated the view (and its rule), but
when I tried to reproduce it to send a usable mail, it evilly refused to
fail and works correctly ever since ;) I fear for the worst, that the
problem comes to life again when it goes from test to live use.
Well, even if it comes to life, I can get back to calling that function; the
problem is, that if I wish to JOIN with table x using this field of the
view, the planner makes a seq scan on table x, unlike when I write the
function body in the view definition.
Since one produces 17sec and the other 0.11sec as total time for the same
query, I think the planner doesn't deliberately choose seq scan; it can't
choose index scan. So, I helped it to see what I want in depth, and that
caused the error -- I repeat, it caused the error but once.
G.
--
while (!asleep()) sheep++;
---------------------------- cut here ------------------------------
----- Original Message -----
From: "Hegyvari Krisztian" <Hegyvari.Krisztian@ardents.hu>
Sent: Tuesday, December 10, 2002 5:12 PM
Dear Gabor,
I am far from being a guru, so I typed "ExecEvalExpr: unknown expression
type 108"
into google and it came up with some seemingly useful hits.