winterkoninkje: shadowcrane (clean) (Default)

Life’s been really hectic lately, but I’ve been getting (slowly) back into working on my Haskell packages. In particular, since the switch from darcs to github I’ve started getting more comments and feature requests, which is nice. Over the next half-year or so, here’s what I’ll be up to in my free time between work on the dissertation and work on Hakaru:

containers — I’ve been appointed one of the new co-maintainers of our favorite venerable library. I prolly won’t be doing any major work until autumn (as mentioned when I was appointed), but I’ve had a number of conversations with David Feuer about where to take things in terms of cleaning up some old maintenance cruft.

bytestring-trie — A few years back I started reimplementing my tries to use Bagwell’s Array Mapped Tries in lieu of Okasaki’s Big-Endian Patricia Tries, but then got stalled because life. I’ve started up on it again, and it’s just about ready to be released after a few more tweaks. Also, now that I’m working on it again I can finally clear out the backlog of API requests (sorry folks!).

exact-combinatorics — A user recently pointed me towards a new fast implementation of factorial making waves lately. It’s not clear just yet whether it’ll be faster than the current implementation, but should be easy enough to get going and run some benchmarks.

unification-fd — This one isn’t hacking so much as dissemination. I have a backlog of queries about why things are the way they are, which I need to address; and I’ve been meaning to continue the tutorial about how to use this library for your unification needs.

logfloat — We’ve been using this a lot in Hakaru, and there are a few performance tweaks I think I can add. The main optimization area is trying to minimize the conditionals for detecting edge cases. The biggest issue has just been coming up with some decent benchmarks. The problem, of course, is that most programs making use of logfloats do a lot of other work too so it can be tricky to detect the actual effect of changes. I think this is something Hakaru can help a lot with since it makes it easy to construct all sorts of new models.

June 2017

S M T W T F S
    123
45678910
11121314151617
18192021 222324
252627282930 

Tags

Page generated 29 Jun 2017 05:24 am
Powered by Dreamwidth Studios