That's really cool. Do you have any performance benchmark between TransactionalMap and java.util.concurrent.ConcurrentHashMap? When should i use these collections instead of java.util.concurrent.* collections?
2013/8/18 Ivan Koblik <ivankob...@gmail.com> > Hi All, > > Almost 4 years ago I developed STM with semantic concurrency control for > my project at CERN. Main feature of this STM is the TransactionalMap that > lets you merge concurrent changes. Library is heavily tested and is very > stable. It has been used in production for the past 2 years. > > Recently I released it on GitHub: > http://entwined.koblik.ch (will redirect to > https://github.com/CERN-BE/Entwined-STM) > > Since Entwined STM was designed to be used from Java I wrote a simple > facade for it in Clojure that you can load with > > (require '[cern.entwined.core :as stm]) > > Entwined STM operates on a memory with a fixed structure, meaning that you > have to define what and how many collections you want to have in your STM > and this can't be changed after construction. To construct memory with 1 > transactional map and 1 transactional queue run this: > > (def memory (stm/create-memory :map (stm/create-map) :queue > (stm/create-queue))) > > It's impossible to access transactional entities outside of a transaction, > to run a transaction you can use "intrans" macro > > (stm/intrans memory data (-> data :map (.put :key1 "value1")) true) > > (stm/intrans memory data (-> data :map (.get :key1))) ;-> "value1" > > First line puts [:key1 "value1"] pair into the map. True at the end of the > body tells the memory to commit this transaction. intrans will initiate > commit if body returns truthy value. Second line just shows that the change > has been committed. > > A couple more words on the implementation: I used HashMap to implement the > TransactionalMap, I copy the backing map for every transaction which may be > expensive for some scenarios. Obvious solution would be to use Clojure's > persistent map. Commits are eventually serialized and protected with a > single lock. If you take a look at the Java source you'll see that > Transaction interface has a second method "committed" that is called when > commit is being done. I use this method to write to the hardware knowing > that execution order of committed callbacks is the same as the commit order. > > I would greatly appreciate any feedback and suggestions. If you have any > questions don't hesitate to ask here or email me directly. Documentation is > still somewhat lacking and I'd be interested to know which parts of it > should be improved first. > > Cheers, > Ivan. > > -- > -- > 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 > --- > You received this message because you are subscribed to the Google Groups > "Clojure" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to clojure+unsubscr...@googlegroups.com. > For more options, visit https://groups.google.com/groups/opt_out. > -- 庄晓丹 Email: killme2...@gmail.com xzhu...@avos.com Site: http://fnil.net Twitter: @killme2008 -- -- 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 --- You received this message because you are subscribed to the Google Groups "Clojure" group. To unsubscribe from this group and stop receiving emails from it, send an email to clojure+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/groups/opt_out.
