"The fact that ast combines strings and does not maintain the original syntax of the document means that original syntax information is lost and a writer walking the abstract syntax tree is unable to reproduce the original syntax of the document."

You should take a look at the RedBaron project which implements a Python parser, provide a high-level representation of the code *and* allow to regenerates the Python code without loosing the indentation, comment, or anything else.

https://redbaron.readthedocs.org/en/latest/

"To achieve this, [RedBaron] is based on Baron a lossless AST for Python that guarantees the operation
fst_to_code(code_to_fst(source_code)) == source_code."