Hi Sean, Sounds like you have greater ambitions than simply supporting transients. Please feel free to disregard any suggestions in the project template and make the *you* would like to implement over the summer. Please post it here or on Melange then we can discuss further.
Thanks! Ambrose On Wed, Mar 25, 2015 at 3:55 PM, Sean Laguna <sean.lag...@gmail.com> wrote: > Hi, > > I am a third year computer science PhD student at the University of > Chicago, and am interested in submitting a proposal for the typed > transients project. I am very interested, in general, about presistency, > transience, and the interaction between the two models for operating on > data. Persistency is actually influence some of my current work on > eliminating race conditions on data structures common in scientific > computing (ghost nodes and other regions allocated to overlapping > processors in distributed memory programs, etc). I've poked around in the > Clojure source code to get some inspiration for my own implementations > (I've done some work on that in C++ and in Nim), and of course have read > the canonical set of blog posts > <http://hypirion.com/musings/understanding-persistent-vector-pt-1> on the > implementation of persistent vectors in Clojure. (Which, if I recall > correctly, Rich Hickey actually cited in on of his papers on Clojure!) > > I would be happy to work on typing transient data structures in > core.clojure, and to work out a means of typing transient that interacts > efficiently and elegantly with other types, especially persistent data > structures. I'm working on a proposal for this now and can send it in a > couple hours, but I wonder if there's a good way of identifying the lowest > hanging fruit for typing a crucial transient data structure in the core > Clojure codebase where I could do a sort of trial run? It would help > immensely I think with coming up with a laundry list of tasks for > performing this, instead of it being more of a casual perusal of the code. > > One endgame for this work that I'm interested in is automatic > parallelization, which Clojure's persistent data structures, in my opinion, > do inspirationally well. Something I'm very interested in is applying the > shared memory concurrency model to a distributed memory environment. > Persistency has natural distributed memory implications with regard to > especially parallelism, because modifications to the "same" value that are > made, at the same logical time, in different, distributed places in memory, > can potentially be reconciled when the values are eventually synced > (perhaps even lazily). Transients could be used in the distributed > locations, and syncing could be done using persistent logic. My thoughts on > this are a bit more fleshed out than as presented here, but I see real > potential for nearly free (from a syntactic point of view) distributed > memory parallelism. I've read about avout <http://avout.io> but I'm a bit > disappointed by their somewhat unleveraged use of persistency and > transience, and about their seeming lack of support for data structures. > > I know that this endgame is a bit separate from the goal of typing > transients, but I believe that the extra information gained from typing > transients could lend itself to a more painless implementation of something > like the above. So, I have a few questions: > > 1. would a very simple distributed memory parallelism implementation > that relies on persistence and transients be desirable as a test-case for > this project? The desire for distributed memory parallelism could itself be > inferred through a type annotation. > 2. would a project that leverages persistent/transients for race-free > distributed memory parallelism be desirable in general, separate from this > project? > - do you agree that there may be some viability here? > 3. for this particular project proposal: is there a simple example of > an instance of a transient data structure in the core Clojure codebase upon > which I can base an itemized procedure for carrying out this proposal? > 4. is it within the scope of this project to also type persistency > that may be missing in the core Clojure codebase, or to convert > non-transient data stuctures that may benefit from transience? > > Thanks! Hope to hear back soon (and sorry for the late correspondence > about this)! > Sean Laguna > > -- > 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/d/optout. > -- 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/d/optout.
