Most of the basic features of Erie are working, but I am still missing let bindings. I looked for inspiration in the syntax of other Lisps and found a couple of options. Common Lisp and Racket are techincally the same because Racket treats [] as (), but I included both because Racket often uses [] to make certain forms stand out.

None of the styles strike me as particularly better. Clojure has fewer characters, but I like the way CL and Racket group the binding and the value more explicitly.

Clojure

(let [a 1 b 2]
  (+ a b))

Common Lisp

(let ((a 1) (b 2))
  (+ a b))

Racket

(let ([a 1]
      [b 2])
  (+ a b))

Regardless of the syntax, implmenting let should be straight forward. Erlang Abstract Format supports match which is very similar. In Elixir

x = 42

is represented as

{:match, 1, {:var, 1, :x}, {:integer, 1, 42}}

in EAF and could be used to represent let bindings.

But before implementing that, I found this answer on StackOverflow and it revealed a new way to look at let. To sum it up, let doesn’t need to be implemented in the compiler. It can be a macro that converts its forms into a tree of nested lambdas. (Samples are using the currently proposed Erie let syntax.)

(let
  [a 1]
  [b 2]
  (+ a b))

is equivalent to

((lambda [a]
  ((lambda [b]
    (+ a b)) 2)) 1)

So instead of adding let as a feature in the compiler, I’ll save the syntax decision for when macros are implemented.