Codetalker has just undergone major revision! :D

The goal of code talker is to allow for speedy development of parsers + translators without compromizing performance or flexibility.

Features:

• Completely python-based grammar definitions [example grammar]
• Fast (cythonized) tokenizing and parsing

...what more do you need?

Here's the process:

tokenize

produce a list of tokens

If you use the builtin tokens, you can get full c performance, and if you need a bit more flexibility, you can define your own token - either based on ReToken or StringToken

parse

produce a ParseTree

The parse tree corresponds exactly to your rules + original tokens; calling str(tree) returns the exact orignal code. Including whitespace, comments, etc. This step is perfect of you want to make some automated modifications to your code (say, prettyfication), but don't want to completely throw out your whitespace and comments.

Abstract Syntax Tree

parsetree -> ast http://docs.python.org/library/ast.html

An AST is used if you only care about the syntax -- whitespace, etc. doesn't matter. This the case during compilation or in some cases introspection. I've modeled Codetalker's AST implementation after that of python. Codetalker does the ParseTree -> AST conversion for you; you just tell it how to populate your tree, base on a given node's children.

Translate

Once you get the AST, you want to do something with it, right? Most often it's "traverse the tree and do something with each node, depending on it's type". Here's where the Translator class comes in. It provied a nice easy interface to systematically translate an AST into whatever you want. Here's an example of creating and filling out a Translator.

For more info, check out my announcing blog post: Announcing: CodeTalker.

Here's the JSON grammar:

# some custom tokens
class SYMBOL(ReToken):
    rx = re.compile('[{},[\\]:]')

class TFN(ReToken):
    rx = re.compile('true|false|null')

# rules (value is the start rule)
def value(rule):
    rule | dict_ | list_ | STRING | TFN | NUMBER
    rule.pass_single = True

def dict_(rule):
    rule | ('{', [commas((STRING, ':', value))], '}')
    rule.astAttrs = {'keys': STRING, 'values': value}
dict_.astName = 'Dict'

def list_(rule):
    rule | ('[', [commas(_or(dict_, list_, STRING, TFN, NUMBER))], ']')
    rule.astAttrs = {'values': [dict_, list_, STRING, TFN, NUMBER]}
list_.astName = 'List'

grammar = Grammar(start=value,
                tokens=[STRING, NUMBER, NEWLINE, WHITE, SYMBOL, TFN],
                ignore=[WHITE, NEWLINE],              # we don't care about whitespace...
                ast_tokens=[STRING, TFN, NUMBER])     # tokens we want picked up in the Abstract Syntax Tree

• modify codetalker to allow for streamed input