On Mon, 27 Feb 2006, Nikolay Samokhvalov wrote: > On 2/27/06, Bruno Wolff III <bruno@xxxxxxxx> wrote: > The alternatives to distinct on are painful. They are generally both harder > to read and run slower. > >'DISTINCT ON' is evil constuction, because (w/o any 'ORDER BY') it >produses unpredictable result, as 'ORDER BY random()' does. And so does UNION in the standard under some circumstances (look at anywhere in the spec that a query expression is possibly non-deterministic), so I think that's a weak argument.
