On Mon, Jul 13, 2009 at 1:36 PM, Daniel<dan.in.a.bot...@gmail.com> wrote: > On Mon, Jul 13, 2009 at 7:55 AM, Mark Volkmann<r.mark.volkm...@gmail.com> > wrote: >> >> On Sun, Jul 12, 2009 at 7:07 PM, Vagif Verdi<vagif.ve...@gmail.com> wrote: >>> >>> Potentially interesting library for clojurians. Java STM >>> implementation: http://www.deucestm.org/ >> >> As best I can tell, this is yet another "on your honor" STM >> implementation. What I mean is that as long as you use the library >> correctly in every place where it is needed, everything works fine. >> Forgot to use it somewhere in an application where it was needed and >> all bets are off. That's one thing I really like about Clojure. You >> can't get away with forgetting to update Refs inside a transaction. If >> you do, you'll be reminded with an exception. >> >> Is there another STM implementation that enforces its use like this? >> > > Well, I'd say that's a property of the language, not the library. You > could use pretty much any STM library in Java if you can make sure > that the executed bytecode has the appropriate calls in place. Have a > look at Terracotta (http://terracotta.org/). It's not an STM in the > sense of intra-JVM transactions, but inter-JVM transactions (they call > it network attached memory), and it's implemented as a bytecode weaver > (meaning you either transform your compiled class files with an > additional step after compilation to add the library specific > instructions, or you do the same thing at loadtime when loading your > code through a special classloader). Terracotta's approach is also "on > your honor" in the sense that it doesn't force you to declare all > variables in TC terms, but the approach could be extended to throw > (compile time) exceptions on undeclared state variables. > > By the way, I would take a good look at Terracotta if I ever wanted to > extend Clojures STM to work across JVMs. You probably would have to > think a bit harder about how you integrate the TC transactions with > the ones from Clojure, since TC transactions are remote calls and > therefore a lot slower, but I reckon it shouldn't be too hard. TC has > enough tools to inspect how your transactions perform.
Oh, and just to complete the picture for intra-JVM transactions on shared state (clustering), besides Terracotta there's also Hazelcast (http://www.hazelcast.com/), and to a lesser degree (different focus) ZooKeeper (http://hadoop.apache.org/zookeeper/) and Shoal (https://shoal.dev.java.net/). But you'll have to check what the libraries are aimed at, since the implementations are designed for certain usecases. And all of these are implemented and used as traditional libraries as opposed to Terracotta. --~--~---------~--~----~------------~-------~--~----~ 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 -~----------~----~----~----~------~----~------~--~---
