William Stein wrote: > On Fri, Jan 30, 2009 at 4:23 AM, Alasdair <amc...@gmail.com> wrote: >> I was recently experimenting with iterators over permutations, and >> using them to find if a graph was Hamiltonian. This was done by brute >> force - no clever tricks - simply by trying every possible permutation >> of vertices and seeing if each was a cycle. >> >> Now there seem to be (at least) two ways of iteration over >> permutations: >> >> v=Permutations(range(n)) >> vp=v.first() >> ...more code here... >> for i in range(factorial(n)): >> vp=v.next(vp) >> >> and >> >> v=(p for p in Permutations(range(n))) >> ...more code here... >> for i in range(factorial(n)): >> vp=v.next() >> >> Now the second is much much faster than the first. On my rather >> elderly laptop, with n=7, the second code takes less then 3 seconds, >> and the first code takes nearly 82 seconds! >> >> But here's the thing: when I tried to compile my function (as a spyx >> file), there was an error converting Pyrex to C when using the second >> iteration; I was stuck with the slower first method (and importing >> Permutations from sage.combinat.permutation). >> >> So how can I obtain a fast iteration over permutations in compiled >> code? > > Cython doesn't support closures, so you can use the second. Yes, > v=(p for p in Permutations(range(n))) is a closure. > > Use this instead of either of your loops: > > v=Permutations(range(n)).__iter__() > ... > for i in range(factorial(n)): > vp = v.next()
Is there a reason you didn't use the less scary alternative below, other than avoiding one function call? v = iter(Permutations(range(n))) Jason --~--~---------~--~----~------------~-------~--~----~ To post to this group, send email to sage-support@googlegroups.com To unsubscribe from this group, send email to sage-support-unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-support URLs: http://www.sagemath.org -~----------~----~----~----~------~----~------~--~---
