Link o' the day: Matt Might on parsing with derivatives
Matthew Might gave this talk at Stanford describing his and his student's work on parsing with derivatives. It was much easier to follow for me than their paper. [Edit: I was originally speaking of their original ArXiv paper. The version I linked to here is th Functional Pearl, which is significantly better. Yay, a victory for peer review! The video is better still, though.] Also, in the first few minutes he motivates the work very well, focusing on open or incremental parsers for network protocols, syntax highlighting, and the zillions of organically grown configuration file formats. I expect the picture is not quite as rosy as he implies, particularly in terms of the cpu and memory available for parsing, and that the current state-of-the-art is not quite as bad.
This discussion examines the disadvantages of current parsing techniques including how they are a mismatch for the normal developers. And in this term a final solution comes in hand of developers known as parsing with derivatives.