Jubil: Hello, Ray

Those who know me know that I have a stack of projects THIS high, and nothing typically ever comes from them. So, uhhh, when something finally happens it‘s kind of neat!

One of the first times I tried to pick up Rust, I did so by writing a little ray tracer. It only supported spheres, and it supported a flat shader. I mean, to say it was simple is an understatement. Last week, while traveling for work, some of my free time was devoted to an old project, Jubil, that I suddenly had a whim to work on, again. It got to the point where I needed a “real” program to try it out with. That’s how the Rust ray tracer fits into the story.

Here‘s ray.jbl:

(define vec 3 list)

(test "vector" 0 0 0 vec
               length 3 =)

;; (origin dir -- ray)
(define ray 2 list)

(test "ray" 0 0 0 vec 1 1 1 vec ray
            length 2 =)

(define origin [0.0 0.0 0.0])

;; ( vec r -- sphere )
(define sphere 2 list)

(test "ray" 0 0 0 vec 2 sphere
            length 2 =)


;; ( vec scale -- vec )
(define vecscale 'f let
                 (f *.) map)

;; ( vec vec -- vec )
(define vecadd 'a let 'b let
             a 0 at b 0 at +.
             a 1 at b 1 at +.
             a 2 at b 2 at +.
             3 list)

;; ( vec vec -- vec )
(define vecsub 'a let 'b let
             a 0 at b 0 at -.
             a 1 at b 1 at -.
             a 2 at b 2 at -.
             3 list)

;; ( vec vec -- float )
(define vecdot 'a let 'b let
             a 0 at b 0 at *.
             a 1 at b 1 at *.
             a 2 at b 2 at *.
             +. +.)

(test 0.0 0.0 0.0 vec 0.0 0.0 0.0 vec vecdot)

;; ( vec -- float )
(define vecmag 'a let
              a 0 at dup *.
              a 1 at dup *.
              a 2 at dup *.
              +. +.
              math.sqrt)


;; ( sphere ray -- result(vec) )
(define intersect-sphere
   'r let 
   's let
    r 0 at 'origin let
    r 1 at 'dir let  
    s 0 at 'center let
    s 1 at 'radius let

    radius dup *. 'rad2 let
    
    origin center vecsub 'w let
    dir dup vecdot 'a let
    
    w 2.0 vecscale dir vecdot 'b let
    w dup vecdot rad2 -. 'c let
    b dup *. 4.0 a c *. *. -. 'd let
    (cond (d 0.0 >=.) (0.0 b -. 2.0 a *. /. 't let
                      dir t vecscale 'rs let
                      origin rs vecadd ok)
           true none))

(define spheres [
          ,@(origin 1.0 sphere)
          ,@([-3.0 0.0 0.0] 1.0 sphere)
          ,@([3.0 0.0 0.0] 1.0 sphere)
        ])

(define width 320)
(define width/2 ,@(width 2 / int->float))

(define height 240)
(define height/2 ,@(height 2 / int->float))

(define total-pixels width height *)

;; takes a pixel and turns it into a x,y coordinate based  on value of width

(define xy width divmod)

;; if vector has any ok? values in it, paint it
;; ( vec -- int )
(define color
     'hue let 'intersections let 
     (cond (intersections (ok?) filter length 0 >) hue
           true 0))
   

(define draw
   (xy 'yy let 'xx let
    width/2 xx int->float -. 'ex let
    height/2 yy int->float -. 'ey let
    0.0 0.0 -5.0 vec 'orig let
    ex ey 100.0 vec 'dir let
    orig dir ray 'r let
    spheres (r intersect-sphere) map
    128 color
    )
   total-pixels times)


"P2\n" io.eprint!
width str io.eprint! " "io.eprint!  height str io.eprint! "\n255\n" io.eprint!
draw total-pixels list (str) map " " str.join io.eprint!

It’s a bit of a mess – I got very excited that something worked!

Running this under the interpreter:

$ time ./jblrun < ray.jbl 2> thing.pgm
7.71s user 0.02s system 99% cpu 7.725 total

This is mostly because it‘s horribly inefficient–we’ll get into why another time.

Jubil also has a compiler, but unfortunately, it doesn‘t yet support the lexical scoping via let that ray heavily uses.

The result here is a NetPBM, grayscale image that looks like this:

It’s wrong, for some reason being rotated, but there‘s something fantastic about a language getting to a point where it can produce something real.

edit: the `xy` function swaps the x, y coordinates. It's fixed by redefining `xy` as `(define xy width divmod swap)`, but correcting a "buggy" celebration post is not something I'm going to do.