Well, I wrote the clojure version after boasting to ninjudd that deftype would let you do it without writing any java at all. I did some simple-minded benchmarking, and while I don't recall the exact numbers his java version is ~10-50% faster. If you look through the git logs, somewhere I have a version that is only 5% slower than his, but when I added efficient disjoin support that slowed down the conj case a bit; we decided that the purity was worth the performance cost. If you want to resurrect that version, let me know and I'll hunt it down.
I also have a version with even more-efficient disjoin but it slowed conjing down to log2(n) instead of log32(n), and that performance hit, while theoretically nil, was very noticeable in practice. If I were using this for some performance-critical task and never disjoined, I'd use ninjudd's version. Otherwise I'd use mine, just because it's easier to build and tweak since it's in clojure. On Jun 28, 12:54 pm, Tassilo Horn <tass...@member.fsf.org> wrote: > Alan Malloy <a...@malloys.org> writes: > > Hi Alan, > > > ArrayMap isn't very performant for large collections. You might like > >https://github.com/flatland/ordered > > in my clojure app, I totally rely on ordered sets. Currently, I use > > https://github.com/ninjudd/ordered-set > > for which I've implemented transient support which is already > incorporated and released. Looking at your code, it basically looks > identical wrt features, except that your implementation is comletely in > clojure (and includes maps) while ninjudd's implementation is mainly > java. > > If you know ninjudd's lib, can you give advice in what use-cases you'd > prefer your own lib? Basically, in my scenario I only conjoin using > `into' but never disjoin, and performance for that is very important. > That's why I've implemented the transient support. But as it turned > out, that didn't improve performance significantly, cause the > calculation of set members is far more expensive than aggregating > them... > > Well, in any case, I think I'll simply try it out at the weekend. > Hopefully, I'm able to get some totally non-empirical benchmark results > from my test suite that I can poste here. > > Bye, > Tassilo -- You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
