I'm not sure to be honest. I suspect most parse trees preserve most of the info but not up to bijection. For instance, Ruby has ways of taking the Ripper output

I'm not sure to be honest. I suspect most parse trees preserve most of the info but not up to bijection. For instance, Ruby has ways of taking the Ripper output and generating source text but there's no guarantee that this output will be precisely the original source text. After all, the parser would need to distinguish between spaces and tabs and store them as tokens to reproduce the text fully.