Re: Group-wise Maximum performance

[Felix Geerinckx]

Jon Richards wrote:

> Well it seemed to speed up a little, but not that much.
> I'm using a BIGINT because the value equates to a java 64 bit long. Is there something more
> approporate I should use? the number is basically ms since 1970.


I was intrigued by this problem and did some testing myself (comparing with max-concat).
You should definitely try:


SELECT
SUBSTRING(MAX(CONCAT(LPAD(time, 20, '0'), data)), 21)
FROM d
GROUP BY name

(which is supposed to be very inefficient ;-)

--
felix