--- Kurt Welgehausen <[EMAIL PROTECTED]> wrote: > > select a from qqq group by b; > > This question was discussed on the list a year or 2 ago. > > The column a in the simple query above is meaningless; it's > an arbitrary value from each group. There are queries, > however, where a non-grouped column is meaningful, such as > a join where the grouping column is a foreign key that > references the primary key of another table.
Although I don't follow your example, I suppose that if someone wanted to select a not-so-random arbitrary member of a group then this GROUP BY extension would do the trick. But I would favor an explicit approach that disallows this non-standard GROUP BY extension, and instead adds a new non-standard, easily understood aggregate function, such as: select arbitrary(a) from qqq group by b; that would give comparable functionality yet still provide the traditional error checking one would expect with GROUP BY SELECT statements. __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com ----------------------------------------------------------------------------- To unsubscribe, send email to [EMAIL PROTECTED] -----------------------------------------------------------------------------
