Hello. If you are reading this, then probably you are taking a compiler course and going to implement your own compiler, and going to read my code. However, the quality of code varies. Hence, I wrote this down as a guide and I wish it, as well as my code, would be helpful.
This project consists of three main directories: compiler/, ir/ and
ir-interpreter. Most of the code is in compiler/. The implemtation of
my IR is in ir/ and I also wrote a naive interpreter for my IR. It is only
a legacy issue that there are three directories.
By frontend, I mean the parsing, semantic analysis and IRgen. The code of this
part is in the corresponding subdirectories of compiler/.
I handcrafted a LL recursive descent parser. Some ad hoc hacks are used to
simplify the implementation. In most cases, you can use parser generators such
as ANTLR to generate your LR parser. The only reason I handcrafted a parser is
that I thought it would take less time. Anyway, if you are going to implement
your own LL parser, maybe you could read the comments in parse/parser.h. I
wrote down the grammar there, which might be helpful.
My implementation of the AST is NOT good. When I wrote that part, I was not familiar with that Java-like design pattern. Therefore, I used public fields instead of make them read-only using getter. You could see my implementation of my IR to understand what I mean here. Except this issue, there are no big problems.
Meanwhile, instead of recording everything in the AST node structure, I assign each node a unique identifier and record the need information such as types in separate containers. I think this design is better since it is much more flexible and elegant (personal opinion).
This part is quite straightforward.
I believe my code is quite clear. On some subtle things such as translating
array constructions, I wrote detailed comments (Cf.
Builder::translateNewArray).
TODO
TODO