Hi everyone, I've been working on a pure binding to libclang that lets you easily analyze any code that the Clang compiler understands. Github: https://github.com/chpatrick/clang-pure Hackage: http://hackage.haskell.org/package/clang-pure Haddock: http://chpatrick.github.io/clang-pure/ libclang documentation: http://clang.llvm.org/doxygen/group__CINDEX.html A short example that prints out the type of every function in a header file (very useful for binding generation): main = do idx <- createIndex tu <- parseTranslationUnit idx "foo.h" [] let root = translationUnitCursor tu funDecs = root ^.. cosmosOf cursorChildrenF . filtered (\c -> cursorKind c == FunctionDecl) . folding (\c -> fmap (\t -> ( cursorSpelling c, typeSpelling t )) (cursorType c)) for_ funDecs $ \(f, t) -> putStrLn $ BS.unpack f ++ " :: " ++ BS.unpack t Or to find the number of gotos: gotoCount = lengthOf (cosmosOf cursorChildrenF . filtered (\c -> cursorKind c == GotoStmt)) root (This uses lens, but the library itself only has one optional lens-style function). Please refer to libclang's good documentation for an explanation of the functions. The only thing to keep in mind is that functions clang_foo are renamed to just foo, functions clang_getBar are renamed to just bar, and types CXBaz are renamed to just Baz in the Haskell bindings. Under the hood, a reference counting system ensures that the C objects are freed at the right time and in the right order, which allows the public API to be simple and pure. The bindings are not yet complete, so if you find something missing please make an issue or better yet send a pull request. :) Patrick