A minimal linear lisp experiment

A-linear-lisp (ALL) is an expression lisp in administrative normal form, all variables are linear and must be used exactly once in every branch of execution.

Functions are non-linear and can only be declared at the top-level. There is no support for closures or higher-order functions.

let-binding:

(let (id value)
    expression)

if branching:

(if id
    then-expression
    else-expression)

function declaration:

(fn (id id1 ... idn)
    expression)

function call:

(id id1 ... idn)

drop - discard a linear value and then continue evaluation of expression:

(drop id expression)

split - linearly break apart a cons cell (id3) into two parts (id1 and id2) and then continue evaluation of expression:

(split (id1 id id3)
    expression)

(== id1 id2)
(< id1 id2)
(> id1 id2)
(+ id1 id2 ...)
(concat id1 id2 ...)
(cons id1 id2)
(clone id)

Strings:

"Hello world"

Integers:

1
124
256

Booleans:

#t
#f

(fn (min a b)
    -- consume `a` making 2 copies which are stored in the cons cell `aa`
    (let (aa (clone a))
        (let (bb (clone b))
            -- consume cons cell `aa`, placing the car into `a1` and the cdr into `a2`.
            -- both `a1` and `a2` contain copies of the value from `a`
            (split (a1 a2 aa)
                (split (b1 b2 bb)
                    -- our comparison consumes both `a1` and `a2`
                    (let (cmp (> a1 b1))
                        (if cmp
                            -- if cmp is true, then drop `a2` and evaluate to `b2`
                            (drop a2 b2)
                            -- if cmp is false, drop `b2` and evaluate to `a2`
                            (drop b2 a2))))))))

(let (q 10)
     (let (z 20)
        (min q z))) -- evaluated to 10

It is important to note that clone, split, and drop are a bit "odd".

This comes down to our interpretation of a strict linear rule; every value must be used once and exactly once in every branch of execution. It is a compile-time error to use a variable more than once in the same branch, likewise it is a compiler time error to not use a variable within a branch.

This is why our min function must first clone the arguments so that we can consume a1 and b1 in our comparison, finally we then consume using drop the value we do not want, and evaluate to the value we want to return.