Question

I’m looking to write a Python import filter or preprocessor for source files that are essentially Python with extra language elements. The goal is to read the source file, parse it to an abstract syntax tree, apply some transforms in order to implement the new parts of the language, and write valid Python source which can then be consumed by CPython. I want to write this thing in Python and am looking for the best parser for the task.

The parser built in to Python is not appropriate because it requires the source files be actual Python, which these will not be. There are tons of parsers (or parser generators) that will work with Python, but it’s hard to tell which is the best for my needs without a whole bunch of research.

In summary, my requirements are:

1. Parser is written in Python or has Python bindings.
2. Comes with a Python grammar that I can tweak, or can easily consume a tweakable Python grammar available elsewhere (such as http://docs.python.org/reference/grammar.html).
3. Can re-serialize the AST after transforming it.
4. Should not be too horrific to work with API-wise.

Any suggestions?

Answer 1

The first thing that comes to mind is lib2to3. It is a complete pure-Python implementation of a Python parser. It reads a Python grammar file and parses Python source files according to this grammar. It offers a great infrastructure for performing AST manipulations and writing back nicely formatted Python code — after all it’s purpose is to transform between two Python-like languages with slightly different grammars.

Unfortunately it’s lacking documentation and doesn’t guarantee a stable interface. There are projects that build on top of lib2to3 nevertheless, and the source code is quite readable. If API stability is an issue, you can just fork it.