This is super neat! i'd encourage anyone who liked this to check out dybvig's thesis[1] - super readable. It's neat to see how you can cleave the runtime system and compile time system. The first two chapters, at least, are pretty cool.

[1] http://agl.cs.unm.edu/~williams/cs491/three-imp.pdf

I haven't gotten through the whole thing yet but it looks great. The textbook I have now kinda stops at AST generation, so I was assuming the rest would be way more difficult than this. This makes it look manageable!

It's a really good article! Nothing too advanced, but concisely explains a lot of the basic concepts of both a compiler and an interpreter.

May I ask how long it took you to write?

Since this is the sort of post I like, I have taken to doing technical posts on a topic myself. The thing I have found is just how time consuming it is. Maybe I'm just slow at this.

Great write-up, thank you! My one suggestion is that you drop the enum meta-programming; it is a distraction. Instead do what you yourself propose: use a simple class with int attributes.

I went to the same class and I just wrote a bytecode interpreter for work. This is a nice follow up I might contact you when I start work again (still on PTO).

This is an amazingly comprehensive writeup! Very well done. Do you intend on implementing a JIT or an optimizing compiler pass? I'd be interested in reading those posts if you eventually write them.

For part 2, try porting it to rpython and see how hard it is to turn it into a full JIT.

It's extremely primitive. Far better tutorials and bytecide designs exist, much better than those perl-, php-, python- alike bytecodes. Unfortunately only such primitive designs get written and used.

SICP e.g. starts with this and then goes into its improvements, whilst better lisps or lua start with tighter opcode designs, and do proper closures and lambda lifting. python cannot even do proper lexical blocks/functions, the most important part of a interpreter/compiler design.