Personally, I feel the hackishness of GROUP_CONCAT will be fulfilled when it gets a LIMIT clause. It may sound crazy but the idea is that it will make things like the groupwise MAX much more efficient if no time is spent building a large string concatenating all the entries – you could just do LIMIT 1 and save a lot of memory and avoid a warning about group_concat_max_len too.

Roland – good to see you here again, was a bit discouraged to have found your comment over there… Wasn’t stealing though 🙂
Thanks for the median query link – nice one!

here’s another trick that uses the same device to calculate the median:

http://rpbouman.blogspot.com/2007/12/calculating-financial-median-in-mysql.html

“UPDATE: Turns out an anonymous comment to Kristian Nielsen’s post suggested the same solution, about a week ago.”

Yeah, it’s mine. I try to sign those dreadful blog interfaces that require me to get an account, but apparently I forgot that time.

BTW – I made a mistake in that particular example, as I forgot to include a DESC for the ORDER BY (the problem was selecting the last version – not the first)

Roland Bouman
http://rpbouman.blogspot.com/

Not so. It was cheaper on my testing benchmarks. I am careful to say “possible” because I cannot guarantee it will be cheaper to use this index on *any* table in *any* database under *any* circumstance, is all.

Shlomi

This post presents an alternative which is very cheap with the right index, possibly cheaper than a JOIN alternative.

Shlomi