History

• Simonyi has created his own company Intentional Software but it seemed that they have no software to sell.
• I also found some Open Source projects (Mozart, XL Programming Language, Lava) which are more or less associated with AST-centric programming but non of them satisfied me.

The prototype revealed some buggy behavior which are difficult to solve based on the design of Study 1. This design flaws also prevent to develop complex AST transformations. Therefore I decided to stop developing Study 1. The last feature has been implemented on June 12. My colleague Rahel Luethy told me that JetBrains MPS is out. She also mentioned the article Language Workbenches: The Killer-App for Domain Specific Languages? from Martin Fowler published June 12, 2005.

I tried MPS and found out that rendering of the AST nodes are not very flexible. There is a lot of improvements necessary in order to become as sophisticated as the ASTcentric Study 1 prototype. I talked four hours with Markus Degen, a friend and software developer, about AST-centric programming. I showed him the ASTcentric Study 1 Prototype and MPS. He encouraged me to publish this prototype even though it is only a proof of concepts. I registered ASTcentric project at SourceForge. It has been approved at the same day. Release of ASTcentric Study 1 Prototype. I started from the scratch with a new framework because I got problems with the prototype concerning references between nodes. The aim of the new development was to allow that a node can refer a node from another AST. For that reason I had to think about unique node and AST identifiers. For the AST identifier I took an approach similiar to the concept of unique Java class names. For the new development I concentrated on the general AST structure, on AST renderering (including a sophisticated algorithm for wrapping long lines), AST rendering language based on an AST, Swing viewer, and Eclipse viewing plugin. At the end of October I realized, that it is not a good idea to use the AST name also as an AST identifier. I started my third approach again from the scratch. Some of the code developed in the second half of 2005 has been reused. First, I developed a concept for a unique AST identifier and its mapping onto file names. For the third approach I concentrated first on the AST structure and the basic classes building them. Also saving to the filesystem and reading from them were developed very early. Later on I started developing a language for specifing valid AST structures. This was the first application of the concepts. In May 2006 I started developing an editor. Initially it was only a console-based editor. In June, development for a Swing-based editor as well as an Eclipse plugin started. Here I could reuse code from the second approach. The aim was to develop something similar to XML, XML Schema, and a schema-aware XML editor. The current state of the project is as follows:

• The AST structure (and base classes) are pretty stable.
• The Validation Language is in an advanced state but not finished. The main open question is, how elaborate should it be? Should it be possible to check an AST only syntactically (like XML Schema for XML) or also semantically like a compiler when parsing source code?
• The graphical editors (for Swing and Eclipse) are in a useful alpha- or beta-state including validation and something like "code completion" based on the Validation Language.