Hi, I have been playing around with the Elixir AST (trying to help with an Issue in Credo) and I found two things that I would like to know more about.
The issue was that Credo has an ABC complexity metric check and it has to detect function calls, but the AST (such as the one that Code.string_to_quoted gives) has very little syntatic information, i.e., I expected that terms would be marked with their syntatic category (variables, functions, etc).
I tried reading the Elixir compiler source or the Code and Macro modules to find if there was a way to get that information but I could not find it.
I also failed to find how to maintain a stack using the Macro.traverse/prewalk/postwalk calls (to keep information about the current lexical scope).
The book on Elixir Metaprogramming is wonderful on writing macros, but does anyone know a resource about analyzing the AST, even if that means diving deeper in the compiler and using lower-level representations? I would be immensely grateful for that!
You can do it via the accumulator argument. Honestly though I just encode my own processing as just normal function calls, Elixir’s AST is really simple.
I could push values to the accumulator in the prewalk function, but how can I pop them at the right time? Because for example in the prewalk it will read a node and then its children, and then it will go to its sibling node. How can I tell I’m at the last child so that I can pop information from the stack? For example suppose the stack is tracking nested function calls
Yep, pre means it is run ‘before’ recursing down, post means it is run ‘after’ recursing down, so you will push environment in on pre and pop the environment off in post.
Also, the prewalk and postwalk are just like using traverse but doing nothing in the other one, so prewalk is like using the pre in traverse while doing nothing in post (fn x -> x end) and likewise for the other way.
