Parsing is fun, but assuming a good baseline allocator arena allocation doesn’t get you a whole lot these days (maybe it’s better under a GC environment where t

Parsing is fun, but assuming a good baseline allocator arena allocation doesn’t get you a whole lot these days (maybe it’s better under a GC environment where tear down does not require a tree traversal?), especially if you’re able to lazily build the AST (this is what I did in JSC years ago). I still feel that there’s a lot of practical parsing stuff that isn’t considered in academia which leads to a near constant preference for LALR and LR parser generators despite them being honestly worse than recursive descent in almost every important metric if you want to do anything at all other than producing a completely generic AST for known good data.