This is an experiment to see if it's possible to create a better performance execution model for parsley, the python parser generator library.

This takes the AST generated by parsley/OMeta, and transforms it into a DFSM, which then parses the input data.

The DFSM is represented by Node instances (interp2.matchers.Node), where each Node has a matcher, and success and failure callbacks. If the matcher inside the Node matches the input data, success callbacks are executed; otherwise, failure callbacks. Any number of callbacks is allowed, but typically one of them will be a "go to the next state (next node)" operation (interp2.matchers.setRule)

To run the tests, simply run py.test interp2

• refactor the thing
• needs testing at the compiler level, with examining resulting nodes
• rewrite the compiler as a compiler + optimizer
• add more support for parsley syntax (the uncodumented "@" etc)