On Thu, Feb 26, 2009 at 09:53:18AM -0600, Jay A. Kreibich wrote:

>   I've always wondered about this... someone please correct me if I'm
>   wrong, but my understanding is that there wasn't any difference
>   between a left and right join except for the argument order.  It
>   seems like implementing right joins should be pretty easy by just
>   having the parser reverse the arguments when it builds the parse
>   tree.  I realize that might screw with the optimizer a bit, but is
>   there anything obvious I'm missing?

When you use a right and a left join together you can't do that.
Sometimes you can restructure the queries to make use or two left
joins or a sub-query, but it's often cumbersome and not always the same
thing.

If in general there is a way to turn something like

  select c1,c2,c3 from m left join l on l.mid=m.id right join r on l.rid=r.id

into something conceptually the same that only uses left joins then
perhaps the optimizer could do this for us.

There might be corner cases though were such a transform isn't
possible.  When it's vague like this I defer to people like Igor who's
SQL abilities constantly amaze me.
_______________________________________________
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users