Indeed, I was working on it this weekend as part of my work on Rolling statistics (RollingPercentile.) I need to submit the patch this week so that we can commit the changes - but there is some pending discussion about API and high level design.
I agree that we should also replace the algorithm for Percentile. Would you care to join me? ;-) On 10/15/07, Hanson Char <[EMAIL PROTECTED]> wrote: > > The current implementation of Percentile relies on sorting the > underlying array. There is much faster way like the use of Hoare's > partitioning that would take only linear instead of n*ln(n) time. > > Has such improvement be considered before ? Any reason not to do so ? > > Cheers, > Hanson Char > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > >
