That's an interesting challenge! What the algorithm really does is to first find a permutation that will sort the string slice, and then apply this permutation to the slice. Perhaps it would be possible to return this permutation in some cheap usable format.
On Tuesday, June 13, 2017 at 7:56:54 PM UTC+2, Thomas Bushnell, BSG wrote: > > That's really nice! > > Consider a case where I am sorting a slice of structs, but the underlying > sort is based on a string within the structs. Or, with some other radixable > datatype besides strings. > > Can we figure out an interface (similar to sort.Interface) which would > permit a solution for either or both of these cases? The former should be > fairly simple to do without sacrificing the speedup; the latter might be > trickier. > > Thomas > > > On Tue, Jun 13, 2017 at 6:20 AM Stefan Nilsson <trollerip...@gmail.com > <javascript:>> wrote: > >> It's well known that radix sort can be faster than comparison-based >> methods for string sorting. With that in mind, I wrote this optimized >> version of MSD radix sort: >> >> https://github.com/yourbasic/radix >> >> It's equivalent to sort.Strings in the standard library and on my machine >> it's twice as fast as when sorting King James Bible and four times as fast >> for the 1k benchmark in the sort package. Of course I can't promise similar >> performance in other settings, but Go get it if you like. >> >> Stefan >> >> -- >> You received this message because you are subscribed to the Google Groups >> "golang-nuts" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to golang-nuts...@googlegroups.com <javascript:>. >> For more options, visit https://groups.google.com/d/optout. >> > -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
