Lapyst devlog #001 - Pro and cons of an own IR

Currently, the lapyst language works like the following:

1. Lex and parse the sourcecode into an AST
2. Enhance the AST via scope information and interfered types for (nearly) each node
3. Walk the AST to generate LLVM IR
4. Let LLVM do it's magic, with some LLD sprinkled in so we get a fully linked result
5. Profit!

The problem with this is the second and third steps; not only store we some non-sourcecode related informations inside the AST, but we face problems when we want to transform the AST, being it inefficenties in the AST creation or be it simple loosing the ability to reason about the AST.

I finally took this problem head on in an attempt to finally solve it, and thats what I want to write today about.