archives

another new language, Chomsky hierarchy Type-0

I would like some of you to read my short article on "Ameba, a Calculator and Metalanguage" at the following URL:
http://computer.wikia.com/wiki/Ameba,_a_Calculator_and_Metalanguage

I have tried to find a language like the one I have unveiled by mind experiments. My research has not discovered a language like it, but I am an amateur, I know no one who can proof read my work, and since I am dyslexic this document quality is questionable. I will answer questions, correct and clarify as needed. I did spend at least a day per page proofreading and rewriting, but I just cannot see all the errors.

I am interested in comments about the technology, not particularly about proofreading errors. but the two probably cannot be separated.

I believe it to be a context-sensitive metalanguage.In general, argument syntax depends on runtime state.

Looking for pointers: mixfix error recovery

I'm looking for pointers to ideas, approaches, papers, and implementations of error recovery in mixfix implementations.

As BitC is progressing, we're becoming progressively more reliant on the mixfix processing layer. In particular, it looks like we're about to handle type annotations in the mixfix layer, where the expression e:T has an "expression hole" on the left and a "type hole" on the right. This isn't a problem per se, but this is one of the natural places to insert a parse error recovery point, perhaps along the lines of (pardon the bison-ese):

expr: expr ':' type
expr: error ':' type

When mixfix only handled simple expressions, the consensus in our community was that error handling at the granularity of sub-expressions wasn't very motivated. But as more and more of the language processing is handled by the mixfix layer it becomes important to think about error recovery in that layer.

By error recovery, I mean recovery such that further useful parsing can continue after an error. It's easy enough to print diagnostics when an error occurs; the goal is that each execution of the compiler should identify as many errors in the input unit of compilation as it feasibly can.

Does anybody have pointers to work that has addressed error recovery approaches in mixfix processing?