I think that racket guarantees that no vector has more elements than the size of the largest fixnum (to support optimizations).
Also, Jos: you might want to use time-apply. Robby On Tue, Sep 14, 2010 at 9:56 AM, Stephen Bloch <sbl...@adelphi.edu> wrote: > > On Sep 14, 2010, at 10:37 AM, Jos Koot wrote: > >> The following measurement shows O(n). >> But O(n) = O(C+n) where C may be a big number. > > More relevantly, O(n) is hard to distinguish experimentally from O(n log n). > In particular, all the sizes you seem to tried are well within a machine > word, so I would expect O(n) behavior in that region (for reasons that other > people have already pointed out). > > Big-O notation is about what happens _in the long run_, as you "approach > infinity". Any experimental analysis will only tell you about a finite > region, so it can't confirm or deny any big-O estimate. Of course, if your > "finite region" covers all the problem sizes you will ever realistically want > to solve, then an experimental analysis is actually more informative than a > big-O estimate. > > > > Stephen Bloch > sbl...@adelphi.edu > > _________________________________________________ > For list-related administrative tasks: > http://lists.racket-lang.org/listinfo/users > _________________________________________________ For list-related administrative tasks: http://lists.racket-lang.org/listinfo/users
