[drdude]

During my 2015 Compilers class, it was indeed the worst step and hardest to get right. We spent (my and my team mate) about 60%+ of the time making that parser actually parse. The lexing, tokenizing, etc is nothing compared to it. In another project, I used Antlr to avoid the headaches, but there are definitely better alternatives out there (some one mentioned Sly in the comments).

Right now, if I want to mess with some transformation passes for an existing language, I would just look for LLVM and Clang for C for example (readily made front end, and backend with hackable middle end).

P.S. I am a hobbyist at this stage and you can consider me knowing nothing about LLVM. My advice is based on my friends' (same colleague) who works for nVidia for their CUDA compiler using LLVM.

Another reason for me to be interested in LLVM is that I used JVM languages (Scala, Java) for the past 20+ years and with GraalVM and its polyglot support for LLVM based languages; whose support was added using LLVM.

my rusty two cents (Rust is awesome too ;)).