Gwyn Evans wrote:
This specific issue is covered in the patch, but just for reference,
the 'gotcha' with rownum is that it's applied before any ORDER BY
clause is applied
I know :-). That's why I didn't use ORDER BY in my example :-).
, so in most cases, it's not going to do what you
want! That's why the classic solution is the one Jeremy posted, where
you have to wrap the 'user' query in an "select * from () where rownum
...;" construct.
Yes. But it is also slower. I was even thinking about proposing
splitting the logic to use subquerying only if there is ORDER BY clause.
But thinking it a bit further I realize that in most cases when one
needs some sane pagination it happens only for ordered result... So yes,
I agree with your correction.
