[REBOL] Re: More on expressions
From: sunandadh:aol at: 8-Feb-2003 6:47
Joel:
> Here are the benchmark results:
Thanks for doing all that work, Joel!
I'm glad to see that one of my "solutions" was so slow that it didn't make
the cut into the benchmarks. That's as much as distinction as being the
fastest :-)
Other dimensions for scoring how "good" the various solutions are include:
Range of data types handled:
** Some existing solutions only work with integers
** Some only work with numbers
** Some won't work with pairs (try a: 6x2 b: 2x6 c: 4x4)
** Some work with most Rebol data type -- though they fail on things like a:
true b: false c: false
** There's just one (I think) that'll work with any Rebol data type.
Flexibility in ordering
** All solutions so far use Rebol's built-in logic for ordering. But if we're
trying to find the middle Czech surname from a list, how quickly can the
solution be adapted to Czech collating sequence?
Flexibility in processing the data
* Find the median, but ignore missing values when doing so -- how many of our
solutions return 10 from this data? a: 10 b: none c: 10 or from this: a: none
b: none c: 10
Number of items to compare
** Very few scale easily beyond three items How quickly can the various
solutions be adapted to 13 items? or N where N isn't known until run time?
The fastest solution usually isn't the one that runs longest (if you see what
I mean) before being replace by something more generalised.
Sunanda.
