This morning Rene de Groot presented his thesis work in our Software Technology Colloquium. He has worked on two case studies of embedding domain-specific languages in Java — applying and refining our MetaBorg approach to `concrete syntax for objects’.

In the first case study Rene elaborated the JavaSwul extension of Java with the SWing Userinterface Language. Rather than the usual spaghetti code of Swing method invocations, the language allows you to follow the hierarchical structure of the Swing class hierarchy to create userinterface objects. For example, the following GUI

is produced by this Java program

import javax.swing.*;
import java.awt.*;
public class Test3 {
  public static void main(String[] ps) {
    JFrame frame = frame {
      title = "Welcome!"
      content = panel of border layout {
        center = label { text = "Hello World" }
        south = panel of grid layout {
          row = {
            button { text = "cancel" }
            button { text = "ok"}

The JavaSwul Examples page contains a number of other cool examples including the creation of menus with eventhandlers and gridbags with a semi-visual layout syntax.

The second case study is the extension of Java with concrete syntax for regular expressions. First it provides a syntax for regular expressions and checks regular expressions in a program against that syntax at compile-time. Second because of the explicit syntax extension, there is no need for escaping special characters as has to be done when encoding regular expressions using strings. Finally, the extension provides syntax for defining string rewrite rules and combining those into compound string rewrite operations (analogously to rewriting strategy combinators in Stratego). Here is an example, with some wiki-like rewrite rules:

public String publish(String page) {
  regex body = [/ <body[^>]*?> .* </body> /]

  regex amp = [/ & /] -> [/ &amp; /] ;
  regex lt  = [/ < /] -> [/ &lt;  /] ;
  regex gt  = [/ > /] -> [/ &gt;  /] ;

  regex escape = amp <+ lt <+ gt

  regex noattach  = [/ <a[^>]*?> \s* Attach \s* </a> /]
                 -> [/ <strike> Attach </strike> /] ;
  regex edittopic = [/ %EDITTOPIC% /]
                 -> [/ <a href="${editAction}"><b> Edit </b></a> /] ;
  input ~= one( body
                all( edittopic <+ noattach <+ escape  )
  return input ;

In addition to the syntactic extensions of Java and their definition using transformations, the implementations include an extension of Martin Bravenboer’s typechecker for Java, so that Java programs using the extensions are typechecked before transformation.

A preview of the work is presented in MetaBorg in Action, a paper for the Summer School on Generative and Transformational Techniques in Software Engineering (GTTSE’05) that was held in Braga, Portugal last Summer.

Rene’s thesis is due soon.

Today our sysadmin asked me to setup a new wiki for use in some course. (Due to early wiki enthusiasm I became wiki master in our department. In fact it took quite some time to convince people that not having a single web master was a good idea.) When I set up my first wikis, one for our Software Technology research group and the other for Program-Transformation.Org, it took me quite some time to get everything installed in the right place, customize the configuration file, and such misery. For instance, on the webserver of our department perl was not installed in /usr/bin as was expected by the CGI scripts of TWiki, but rather in /sw/bin. This required patching all scripts each time I upgraded the installation. Also the directories containing the site were mounted on a different path in the webserver than on the user server. Thus, scripts that were invoked via cron job (for example to update statistics) needed a different path configuration than scripts running on the server. As a result, updating the site was not something to look forward to, let alone adding new installations.

Today I didn’t flinch a second when receiving the request to create a new wiki. I just copied the 60 line Nix file of another wiki, changed the name of the wiki, the directories in which to store the wiki data, the port at which to approach the wiki on the server, and customized a few other options. Then I called the install script, and there I had a new wiki set up and running. Well, actually I had to ask sysadmin to open the new port, but then it was just there.

This is all thanks to service configuration management with Nix. In this paper that Eelco Dolstra will present tomorrow morning at SCM-12 in Lisbon (Portugal), we show that the management of services requires the integrated management of software and configuration. By treating software configurations just like any other software components, rather than as state managed in global directories, one can uniquely and reproducibly describe a specific instance of a service. We have used this to succesfully deploy quite a few services already, including a subversion version management server, a Jira issue tracking server, a buildfarm and automatic release management system. The description of the machine specific parts of the deployment of these services is factored out into a small Nix expression, such that I could create a combined subversion/wiki server on my little home machine with the push of a button (well almost).

So if you don’t have a chance to attend Eelco’s talk tomorrow morning, at least have a look at the paper.

The new translation scheme turned out quite well. On Linux it gives a slight performance improvement, on Macs it gives a major performance improvement; speedups of 3 times have been observed. Good reason for a new release:

Stratego/XT 0.16M1 is now available from
This release is the first milestone release (M1) towards the new major release 0.16. (The zero-th milestone was called 0.15.)

The purpose of the 0.16 development is a major overhaul of the language and compiler internals. Release 0.15 introduced the Stratego Core language and the corresponding refactoring and clean up of the compiler. This release features a lot of fixes to the Stratego Core compiler and lots of other small improvements. Noteworthy features include:

  • call(f|s*|t*) is a new language constructs that supports calling strategies by name, i.e., f is a term that is interpreted as the name of the strategy to call. Can be used for callbacks in libraries.
  • checksum strategy gives MD5 checksum of a term
Furthermore, this release already contains some features planned for future milestones:

For 0.16M2: A new translation scheme for implementing choice. Rather than using setjmp/longjmp the scheme now use NULL pointers to indicate the failure of a strategy. This produces a slight performance improvement on Linux and a great performance on Macs, where the setjmp/longjmp feature is quite expensive to use.

For 0.16M3: A good number of improvements to the implementation of dynamic rules. In particular, the compiler now detects overlapping dynamic rules and forbids their use.

While the 0.15 release was still rather experimental, the 0.16M1 release is much improved and is fairly reliable. We are interested in your experiences. Please report any problems such that we can solve these as soon as possible. We plan to have a stable 0.16 release by the end of September.

After deliberating about it for a long time (several years even), I finally implemented a new translation scheme for the Stratego Compiler. The old scheme used the C feature of setjmp/longjmp to deal with failing transformations. This provided the opportunity to go from using C as an assembly language, where an entire Stratego program was compiled to a single C function using gotos for control-flow, to a more idiomatic style of C programs in which each strategy definition was compiled to a C function. The setjmp/longjmp feature elegantly dealt with the notion of a failure by declaring a choice point (with setjmp) and jumping to it from anywhere (with longjmp). However, since choice points are the control-flow mechanism in Stratego, the speed of programs depends heavily on the cost of this feature. On Intel machinery (running Linux) this is not a big issue, but on Apples and Suns (RISC machines) the number of registers saved at each choicepoint is quite expensive; at least that is a theory about possibilities for improving the performance of Stratego programs.

Eelco Dolstra suggested a long time ago to return NULL to indicate failure of a strategy. Indeed, this representation closely matches the formal operational semantics of the language, in which the set of terms is extended with a failure value; exactly the ATerm data-type extended with an extra value (NULL). While conceptually simple, the idea seemed too disruptive to the run-time system, compiler, and native parts of the library to start work on, and there were always plenty of other things to do. (Especially considering the fact that the change does not add one bit of functionality.)

After the recent Stratego Core refactoring and clean up of the intermediate representation and the compiler, and solving lots of outstanding issues, this translation scheme refactoring came into view again. Interestingly, it took less then a week real time to achieve, which included a camping trip, a visit to Philips Research, and reading a couple of theses and articles. So either the problem was never that problematic to start with, or the recent drastic refactorings of the Stratego/XT source tree, configuration, and build system has paid off. Also the change to baseline-based bootstrapping (from self-based bootstrapping) has enormously simplified the process of changing the foundations of the compiler. Finally, the reliance on a solid continous build and release system gives one much more confidence in committing to a new version to bootstrap against. If there is a general lesson here: refactoring and continuous integration pay off.

As for the new translation itself; it is pretty standard fair. Have a look at s2c-ng.str and compare it to the classic s2c.str. The only flaw is that I had to resort to producing code with gotos. Noteworthy about the new version is the use of concrete syntax of Stratego and C almost everywhere, which makes a difference between night and day in maintaining the translation scheme. For example, the following rule defines the translation of the crucial guarded choice construct:

  TranslateStrat(|S,F) :
    |[ s1 < s2 + s3 ]| ->
        ATerm ~id:x = t;
        ~id:F' : t = ~id:x; 
    where <not(Next)> S; new => x; new => F'

Another interesting new feature is the collection of code fragments using dynamic rules, and the synthesis of the target program from these fragments afterwards; in contrast to the old method in which the source program was traversed for each type of fragment `driven by’ the target program.

I’m writing this blog while waiting for the bootstrap build to go through, but from the regression tests for the compiler the implementation seems to work fine. What is not clear yet, is the performance improvement this will bring, if any. Another feature of the current translation scheme that seems to have negative impact on the performance of Stratego programs, especially on Apples and Suns, is the use of nested functions in the target C code. This is a feature of GCC, and therefore fraught with portability and performance problems. The implementation using trampolines also does not go very well with the tendency to forbid executable code on the stack. So this aspect of the translation is next in line to be changed. The basic idea is again simple and the use of fragment collection was partly motivated by this change. More later.

Among the useless things one can do in life, maintaining one or more publication lists ranks high. My tendency to waste time on my publication list probably dates back to my days as a PhD student when I badly needed publications to put on a list. On the other hand, as publications are the measure of achievement in research, more researchers may have this problem.

Anyway, maintaining a list of publications can be quite tedious, in particular if you want to provide multiple views on the publications. For example, a listing with most recent publications first, one providing the most important ones first, one organized by research topic, and finally a separate list for each project. Also your department may require regular submission of lists. On the web version the entries should come with links to the pdf files and/or the webpage of the publisher, but these links should not be displayed in the version for printing, since they are quite useless there.

Being a computer scientist, I elevated the activity of maintaining content to maintaining a program for generating the various lists. This is still a waste of time, of course, but the excuse is that it will save me time in future. Another excuse is that I developed my program as a case study for the transformation language Stratego.

In fact, the bibtex-tools package has emerged over a long time, starting with a syntax definition for BibTeX first written in 1999. It turns out that BibTeX has quite an intricate syntax that is not so easily formalized with a traditional approach based on a separate lexical analyzer and context-free parser. With the scannerless approach of SDF this poses no problems at all.

Also, the use of the Stratego to perform transformations on a structured representation of a BibTeX file is a definite improvement over directly transforming its text representation. Moreover, these transformations can be expressed quite concisely. For example, the following strategy definitions define an inliner for BibTeX that replaces occurrences of string identifiers with their body. (BibTeX allows the definition of strings such as @string{LNCS={Lecture Notes in Computer Science}}, which can then be quoted in entry fields using the identifier, e.g., series = LNCS.)

  bib-inline =  bottomup(try(DeclareInlineString + InlineString + FoldWords))

DeclareInlineString = ?String(_, StringField(key, value)) ; rules( InlineString : Id(key) -> value )

FoldWords : ConcValue(Words(ws1), Words(ws2)) -> Words((ws1, ws2))

After having developed my own set of BibTeX tools using the Stratego transformation language over the last couple of years, I decided to make them into a proper software package that could be used by others, complete with a manual that explains the LaTeX/BibTeX/Hevea techniques used to get a publication list into HTML. The currently availabe version is a pre-release of the first official release 0.2. I’m waiting for a new stable version of Stratego/XT and for some feedback from users before I make the release official.

So if you don’t want to waste time on editing your publication list webpage, but instead want to wast time learnig to use my tools, you now know where to find them.