The book is really great. It starts with a very interesting explanation of lambda calculus and how high-level functional programming language can be tied to it. Chapters 8 and 9 were written by Peter Hancock and cover step by step how to write a polymorphic type-checker (the examples are in Miranda but it's not far from Haskell).
I once had an embarrassing post-concert-in-the-pub conversation with PH in which I confidently asserted that he didn’t understand polymorphism. “Yes I do. I wrote N chapters of a book about it”. I call BS and demand name of book. He names book. I say “but I have that book and it’s by SPJ”. “Ah” says Hank “but the chapters on polymorphism are mine!”
Self plug: Here's a Rust implementation of [one of the Virtual machines described in the book (TIM)](https://github.com/bollu/timi) I worked hard on the UI, and I hope it shows. Exploring the states through the UI gave me deep insight into how the machine works.
<bool> ::= T | F | IsZero <num>
<num> ::= O
<arith> ::= Succ <num> | Pred <num>
Here `<num>` can only ever be `O`, so that's the only valid argument for `IsZero`, `Succ` and `Pred`. I would put `O` into `<arith>`, get rid of `<num>` and replace all references to it with `<arith>`.
That doesn't affect the actual implementation, since (as they say right after) "For simplicity, we merge all them together in a single Term."
The book is really great. It starts with a very interesting explanation of lambda calculus and how high-level functional programming language can be tied to it. Chapters 8 and 9 were written by Peter Hancock and cover step by step how to write a polymorphic type-checker (the examples are in Miranda but it's not far from Haskell).
For a 1987 book, it all aged very well.