APL Interpreter – An implementation of APL, written in Haskell (2024)

Summary

APL InterpreterJanuary 11, 2024Why APL?APL is an array programming language. Its only data type is the (multidimensional) array. While this might seem like a huge limitation, the generality it provides leads to a syntax that is incredibly compact and expressive, which forces the programmer to approach problems at a higher level.I was first drawn to APL by the nature of its syntax: with the exception of user-defined variables, all built-in functions and operators are single unicode symbols. As a result, code looks like this.⍝ My solution to day 7 of Advent of Code '23 c2n ← '23456789TJQKA'∘⍳ classify ← (5-≢⍤∪),(⌈/(+/∘.=⍨)),((5≡∪)×(⌈/c2n)) h2n ← 13⊥classify,(¯1∘+c2n) hands bids ← ↓⍉↑((' '∘≠)⊆⊢)¨ input bids ← ⍎¨ bids ⎕ ← +/ (⍳≢bids) × bids[⍋ h2n¨ hands] modes ← { ⍵≡⍬ : ⍬ ⋄ ∪⍵⌷⍨⊂(⍸⌈/=⊢)+/∘.=⍨⍵ } change_joker ← { (⊃ 'A' ,⍨ modes (⍵~'J'))@(⍸⍵='J') ⊢ ⍵ } c2n ← 'J23456789TQKA'∘⍳ h2n ← 13⊥(classify change_joker),(¯1∘+c2n) ⎕ ← +/ (⍳≢bids) × bids[⍋ h2n¨ hands]As you might expect, learning to write programs like this requires a totally different mind-set. Array programming is similar to functional programming – the primary way to control execution involves composition of functions – but APL tends to encourage the reliance on global properties and sweeping operations rather than low-level recursion.I’d like to think that learning to approach problems this way provides greater insights for programming in general.Why Haskell?I originally planned this project to be a deep-dive into APL, and learning Haskell was more of a side-quest. It ended up the other way around: the hardest aspect of this project, by far, was learning work with Haskell.In all honesty, Haskell probably isn’t an ideal tool for an array-language interpreter: it makes parsing and combination of functions much more elegant, but at the expense of ease of working with state, data structures, and performance.As a result, this project isn’t intended to be especially practically useful, nor to be a replacement for existi...