Peter Eisentraut wrote:
>
> > > playpen=> select ta.a,ta.b,ta.c, (select count (tb.zz) where tb.yy =
> > > ta.a) from tablea ta, tableb tb order by tablea.a;
> > [ produces 80 rows ]
>
> > > playpen=> select ta.a,ta.b,ta.c, (select count (tb.zz) where tb.yy =
> > > ta.a) from tablea ta, tableb tb order by ta.a;
> > [ produces 20 rows ]
>
> > > playpen=> select tablea.a,tablea.b,tablea.c, (select count (tableb.zz)
> > > where tableb.yy = tablea.a) order by tablea.a;
> > [ produces 4 rows ]
>
> Once again, I think that we *really* need to discuss whether implicit
> range table entries in SELECT are a good idea.
What is an "implicit range table entry"?
We invariably get a
> question like this every week and invariably the answer is "if you give a
> table an alias you *must* refer to it by that alias".
Hey, I *did* do that in the second query, and that still produced extra
results. I tried putting the aliases in the inner select too but that
didn't help. In fact the inner select always is 4 in that case. Unless I
only alias tableb in the inner query, and let it get the definition of
tablea from the outer query.
(I'm sure Tom has
> this reply automated by now.) I claim the only thing that buys is
> confusion for very little convenience at the other end.
>
> Stop the madness! :)
>
> --
> Peter Eisentraut Sernanders väg 10:115
> peter_e@gmx.net 75262 Uppsala
> http://yi.org/peter-e/ Sweden