weblogging at the beach

« April 2006 | Main | June 2006 »

May 31, 2006

Compile Your Code!

Here at Untyped Central Dave and I are hacking away like lumberjacks, which is a good thing as the project we're working on is due soon. We're constantly running tests and loading code into the web server, and we've noticed that these processes have been getting slower and slower. It turns out the bottleneck is the time to parse and byte compile our Scheme code. Simply byte compiling the code beforehand has made an incredible difference. Tests that used to take minutes now run in seconds. Two features of PLT Scheme make it really easy to integrate byte compilation into our development process. Firstly, mzc will follow dependencies when given the -k flag. So we just run mzc -k main.ss and all our code is compiled. Also useful is that PLT Scheme does the Right Thing and loads source code if it's newer than byte compiled code, so we don't have to constantly recompile our code. So we can just code away as normal, except every time we take a break we run mzc. Eventually we might write some code to recompile at regular intervals (say, every 10 minutes) but for now it isn't worth the effort.

Posted by Noel at 05:43 PM | Comments (1)

May 24, 2006

Of Interest 24/05/2006

If you're a web developer then you pretty much have to read Brendan Eich's slides on JavaScript 2 / ECMAScript Edition 4. The quick summary: it looks like Scheme with an optional type system, which is a way of saying it looks damn good. All Hail Professor Bloggy McHerman!

Gilad Bracha's rant against continuations and the subsequent followups are interesting and infuriating in equal parts. Once again Prof B comes to the rescue. His pedantic aside is particularly important. I'll also add my own brief rejoinder: I'm not an idiot. Give me all the tools in the toolbox, and I'll make the decision when to use them. In some situations it is appropriate to use cookies (e.g. login), for others encoding continuations in URLs is the right solution (e.g. persistent URLs), and for others server-side continuations are right (e.g. temporary pages such as form validation)

Seems like no-one really knows how much sleep you should get, but a drug is in development that might do away with the need, for short periods of time. When travelling around Spain I found siestas worked really well. I got by better on 2 sleeps a days, and studies of experts have shown the highest performers often take an afternoon nap. Matt's on to something in his tree-house.

Posted by Noel at 05:44 PM | Comments (0)

May 18, 2006

Interesting Stuff 18/05/2006

Amazon is a platform, not a retailer. And an interesting platform at that.

The algorithmization of science is the next step to applying mathematics to problems that have been difficult to model equationally. Humanities students better dust up on their computer science!

Google releases a AJAX framework. Catching up to Yahoo on this one. Does anyone care? Well, pundits say AJAX is no different than eating large quantities of chocolate so I guess a lot of women should.

Posted by Noel at 04:03 PM | Comments (0)

May 17, 2006

How American are Startups?

Dave, living it up at XTech in Amsterdam, IMed me about Paul Graham's latest. A quick search finds that lightening fast typist Suw Charman offers the best synopsis: Xtech 2006: Paul Graham - How American are Startups?.

There are some good points; nothing surprising if you've ever been involved in a startup but it is worth a read. My main gripe: there is no Europe. It is a collection of disparate countries with vastly different governments, values, and so on. Capitalist UK is not socialist Denmark.

That aside, by Paul's criteria the real Cambridge is pretty close to getting it right. It just might be too expensive for real in-the-garage type startups. When paying the rent is a challenge spending time on a risky proposition like a startup seems less attractive.

Posted by Noel at 05:23 PM | Comments (0)

More Dagstuhl Talks

Modules
Matthew Flatt

A module is "the way you share code with other programmers"

Module systems in different languages: Python, Ruby

Mutually dependent modules are the first issue

Ruby example: mutually dependent modules sometimes work (when required from files) but dependent on order of execution. Shows that a module definition is a side-effecting operation. Similarly a module may be extended. Hence a module is a runtime concept.

Ruby modules also function as mixins.

MzScheme modules are not side effects. Runtime order does not matter. However expand time (aka static analysis time) order does. Mutually dependencies disallowed. Use units for mutual dependencies. (New!) Unit/signature modules. A module can be written in a unit or signature language. External linking -- the module no longer decides what it links to (cf internal linking, the usual method).

Scheme48 modules, known as structures. Similar to MzScheme units and ML functors. However allows units to export macros, unlike MzScheme.

Smalltalk modules
Alexandre

Smalltalk 80 has no modules. Distribute changesets. Changesets have no composition.

VisualWorks/Squeak packages track their changes and so can be uninstalled. Packages have prerequisites.

Selector namespaces in Modular Smalltalk. They are not reentrant. Seems to have strange scoping rules.

Classboxes allow re-entrance.

CPAN
Mike Sperber

Pragmatics and deployment.

Tools create boilerplate. Standard format for packaging and creation.

Online distribution (CPAN) and shell for retrieval and installation. Can automatically install dependencies.

Perl code may branch depending on installed modules and versions.

Social features (e.g. number of available modules) make it difficult to change the implementation. For instance, you can only have 1 version installed at a time. Cf PLaneT allows multiple versions. However there are still issues with loading multiple versions at once.

Java's Future
Gilad Bracha

JSR-277 attempting to add a module system for Java. Main interest is deployment, versioning etc.

Initial proposed system: modules are like units -- parameterised.

Actual proposed system: no language changes allowed. Instead embed using reflective API. Turing complete therefore any particular implementation possible (first-class, higher-order, etc.)

Module implementations check for prerequisites at runtime. Just code, so anything is allowed. Lots of state.

My comments: As usual for Java has taken a reasonable idea and turned it into a huge hairball.


Soapbox Session

Context-oriented Programming
Pascal Constanza

Make program change behaviour due to context of use (e.g. personalisation) without making it a huge hairball.

Basically a new type of modularity. Similar to aspects, and OO inheritance / overriding.

F-Script
Philippe Mougin

Open source scripting language for Cocoa / OS X

Unifies OO and array programming

Interactive environment.

Source Code Mining for Latently-Typed Languages Dave Mandelin

Extracting information from source code. Perhaps to find out how to use a library from example code.

Example: what are the types of the arguments in a latently typed language?

[Reflection!]

BabyUML
Trygve Reenskaug

Modularity, again.

My comments: I didn't really get this talk, so my summary is brief and perhaps incorrect.

Ambient Oriented Programming

Ambient resources and volatile connections
Non-blocking communication
Reified communication traces
Reified environmental context
OO sans classes to handle code updates

Converge
Laurence Tratt

Pythonic language with compile time metaprogramming ala Template Haskell.

Converge's role is to host DSLs, and so must be very flexible. In a senese a Lisp with curly braces.

Customisable parsing.

Father Time
Greg Cooper

MzTake
Guillaume Marceau

My comments: I've seen FrTime and MzTake before so I didn't take notes. If they're new to you check them out -- they come as part of PLT Scheme

Posted by Noel at 09:47 AM | Comments (0)

May 08, 2006

Binding! The beer of choice for Computer Science

David van Horn and I were wandering around Frankfurt after Dagstuhl when we came across this sign. It's beer, and it's also one of the major issues in programming language design. Clearly this is the drink of choice from discerning computer scientists!

Posted by Noel at 09:09 AM | Comments (0)

May 04, 2006

Dagstuhl 06181

Due to what I can only assume was a clerical error, I was invited to the Dagstuhl seminar on latently typed languages. I accepted before this error could be rectified, and so it is that this blog post comes to you from room 23 in Dagstuhl.

There have been a number of talks of various topics of interest. Being a good student, I've taken notes, which I intend to put online as time allows. You'll have to fill in a few gaps but I hope they'll give you an idea of what has been discussed. The first talk is below:

Cross Language Runtimes
Wolfgang De Meuter, Marc Feeley, Robby findler, Roel Wuyts et al.

Sapir-Whorf

Ralph Johnson: study at IBM on what characteristics making people learn Smalltalk more easily. Number one characteristic was how many languages you already knew. The more you know the easier it is to pick up new languages.

Matthias F: ppig.org Productivity of programmers is directly dependent on number of languages they know not the number of years of experience.

Cross-languages

Survey cross language systems

Gambit-C (Scheme) FFI to C/C++

Matthias F: Scheme is not a safe language as the spec leaves many things undefined.

ProfessorJ

Robby F: dynamic is like the Any (TST) type for Java. Too much to explain now.

SCM2JS Scheme to Javascript

SHard Scheme to VHDL

Soul Prolog and Smalltalk

etc [other examples]

Linguistic symbiosis

Robby F: sometimes wrappers must stay around to enforce constraints when, for instance, mixing static and latently typed languages.

Andrew Black: copying implementations don't work in parallel systems with mutable state

My comments: There were a lot of questions so the talk ended up being a bit rushed. A lot of time was spent on the survey (and answering questions) which unfortunately left little time to really tease out the interesting issues.

Posted by Noel at 11:38 PM | Comments (0)

The Rumour Starts Here!

Salacious rumour suggests the next version of Javascript will have tail recursion. We can neither confirm nor deny this rumour, but we can rock an old style w00t!

Posted by Noel at 11:01 PM | Comments (0)