On 2015-09-12 10:58 PM, Aurel Wisse wrote: > The fastest solution is actually the temporary table:
Glad you found a faster solution. > Still Richard : How about allowing recursive aggregate queries again ? You say "allow" like it's something that worked and they blocked it out for no good reason. It didn't work, your use case is a single accidental fringe that happened to work by virtue of the MIN() operation's character in the specific query, much in the way that if you throw a handful of seeds on a pavement and one of them happens to roll all the way over to a mossy moist area and found it could germinate, it is not a good reason to promote pavement-seeding - even if you happen to be the observer from the mossy patch. Put another way: It isn't good to allow a convention that will "seem" to do a certain thing or work a certain way, but then give a completely wrong answer in 99% of cases - all because 1% happens to be correct. MIN() actually has two implementations, one aggregate and one as a simple function, might be one could allow MIN() and MAX() if they always produced the correct results... but then we need test-cases to ensure they always give correct results, we need to maintain code that caters for it etc. and all that to achieve a prize which is handled better and faster by a temp table in the end.
