Lambda the Ultimate

Interesting project.

There seems to be intrinsic appeal in making a language its own meta-language, and there are certainly advantages to this approach (e.g. you only need to know one language to write both programs and meta-programs).

MetaC is a pretty good illustration of the disadvantages of this approach:

- C isn't a very convenient language for expressing the kind of data structure-rich access and manipulation that you are likely to want for sophisticated meta-programming. In order to cope with this, MetaC seems to have exploited undefined corners of the language (e.g. out-of-bounds array accesses) to add more "convenient" behavior (e.g. growing the array to accommodate the out-of-bounds access). This just screams "kludge".
- Implementing a full C compiler/interpreter is a lot of work compared to the equivalent task for most other languages. Sure, he can probably reuse a lot of the code from the C program analyzer. But even taking that into account, you're probably still looking at more than an order of magnitude extra work compared to writing a simple Scheme interpreter.

In my opinion the real value of something like MetaC is in the C program analyzer and the API for accessing and manipulating programs in a structured and convenient way. Once he releases the code someone should tear out the MetaC language interpreter parts, and bolt on a simple interpreter as a replacement. :)

This reminds me vaguely of `C which I don't think was ever mentioned on LtU (but I haven't checked). Another Dawson Engler (Xok/metal) related thing.

They seem to have moved the page to:

http://www.rcs.ei.tum.de/research/MetaC/index_html

Some other useful links for MetaC:

A paper.

Maier-Komor's MetaC site.