The Mendel Project

Overview

The Mendel project is my attempt to design a system for writing software that doesn't depend on programming languages. Instead, it will allow programmers to manipulate software in terms of abstractions rather than synax. The Mendel project is named after Gregor Mendel.

Current Status

So far it's just me and some Python scripts, so don't get too excited.

My current task is to develop and implement an object model that will be used to represent programs. Since premature optimization is the root of all evil, I'm trying to come up with something as simple and general as possible in order to make things as flexible as possible. Here are my latest thoughts on the subject.

Development Roadmap for Version 1.0

The first stage of development must follow a fairly linear sequence of steps. The end of this sequence will be version 1.0.

• Develop and implement an object model for Mendel. (Done?)
• Define a set of objects to represent Python programs.
• Add functionality to tranlslate Python programs into Mendel objects. Fortunately that hard part (parsing Python) is already available as an API of the Python runtime system. This will be a good proof of concept for the object model and will provide a source of non-trivial objects for testing.
• Make the Mendel-ized Python programs capable of executing themselves. This sounds hard, but it just involves having the Mendel objects "un-parse" themselves into Python code and call the Python interpreter. At this point the Mendel system is just a very baroque Python interpreter that works by parsing Python code, unparsing it, and then passing the result to the real interepreter.
• Implement persistence for Mendel objects. Make sure it is CVS-friendly. Add a UI for parsing Python code and saving the result as a file and a UI for executing a set of Mendel objects stored in a file.
• Use the parsing UI to translate the Mendel code base into a file containing Mendel objects. Use the execution UI on the resulting file to re-parse the code base into a different file. If the files are the same, the Mendel system can run itself.
• Write a GUI for editing Mendel objects. This will use the wxWindows library for platform independence and because it has a Python binding available.
• Write a Python parser that preserves comments.
• When the GUI is mature enough, switch all development over to using Mendel objects rather than Python source. This can be done in a single step using the comment-preserving parser. This is version 1.0!

Future Directions

After version 1.0 there are several different features that can be added more or less independently. This is where I could use some partners.

• Define a "core language" of Mendel objects that can be translated to C or C++ as well as translations for Mendel-ized Python objects into the core language. This will end the reliance on the Python runtime system.
• Refine the GUI. Make it easier to use than a text editor.
• Create Mendel objects for non-Python language constructs capable of translating themselves into the core langauge or Python constructs.
• Add parsers for languages other than Python.
• Integrate with the Mozart project.
• Add platform-specific extensions.
• Add a native revision control system. This will pave the way to representing Mendel objects in space-efficient binary formats and database schemas rather than being restricted to CVS-friendly text files.

Links

Here are some links that may or may not be relevant. I haven't really checked them out yet.

1. Mozart
2. the Cecil language; interesting object semantics
3. Eidola
4. Fabrik
5. Cube
6. dmoz.org: visual programming languages
7. dmoz.org: metaprogramming

Last update: $Date: 2002/09/10 15:57:54 $