Fri, 06 Jun 2008

FP-Syd #4.

Last night was the 4th meeting of FP-Syd, the Sydney Functional Programming group. First up I'd like to thank Shane, James and the other Googlers who organised the use of one of Google's Sydney meeting rooms and laid on drinks and a bunch of tasty snacks. Another big thanks goes to our speakers for the evening and the 32 (!!!) people who showed up.

Unfortunately one of our planned speakers for the night had to drop out at the last minute. This meant that as people arrived I was asking them if anyone had a short talk they could give at very short notice. The thing that impressed me the most was that we got two offers.

First up we had Sean Seefried talk about using Ocaml at Nicta to write Nicta's Goanna static type checking tool for C and C++ code. It was interesting that they were using an existing commercial C and C++ parser and the converting the C abstract syntax tree (AST) produced to an Ocaml AST where they work on it further. Also interesting was that the chose Ocaml because it was a good language for writing compilers as well as having reasonably good and highly predictable performance. This was an excellent talk considering that Sean had about 10 minutes to prepare. Thanks Sean.

The next speaker was Scott Kilpatrick who also had about 10 minutes to prepare as well as being our first international speaker. Scott has just recently completed two B.S. degrees (one in Computer Science, the other in Mathematics) from the University of Texas at Austin and will be starting a Masters there in September. He's is currently interning at Google here in Sydney working on the Google Maps project.

Scott gave us a very interesting introduction to SUN's Fortress language. Scott is working on implementing the type checking system for the Fortress compiler rather than being a dedicated user of the language. He was however able to give us a brief overview of the language, which is designed to replace Fortran in high performance computation tasks. Fortress seems to have a huge array of interesting features and is definitely something to keep an eye on. Thanks Scott.

Finally we had Tim Docker speaking about writing a Tuple Server in Haskell using the State Transactional Memory monads in concurrent Haskell. Tim did a great job explaining what Tuple Servers are and what they are used for where he works at Maquarie Bank. He then talked about the monads in Haskell, in particular the STM monad and its use in his tuple space server. It was a great talk and I mid way through it, I realised that a tuple space server is a nice solution to an architectural problem I have. Thanks Tim.

ddc logo

After the talks we retired to the Redoak hotel for a few beers and a chat. Since a couple of people had missed the previous meeting, Ben Lippmeier, was kind enough to pull out his laptop and give his talk on "The Disciplined Disciple Compiler" again. Interestingly for me, he went into a little more detail on areas that he glossed over in the last meeting. DDC is a really interesting piece of work, being a strictly evaluated dialect of Haskell, with optional lazy evaluation (far more elegant than the way Ocaml does it) and with side effects and destructive updates looked after in the type system. The ideas behind DDC are really compelling. I look forward to seeing it progress. Thanks Ben.

Posted at: 22:18 | Category: FP-Syd | Permalink