Caps at beginning of curve

[diyaayay]

Resolves #6816
Work in progress.

Changes:

-Handles add cap at the start of the stroke if not connected

Screenshots of the change:

sketch:

function setup() {
  createCanvas(400, 400, WEBGL)
}

function draw() {
  background(255)
  stroke(0)
  strokeWeight(20)
  beginShape()
  vertex(-40, -40)
  bezierVertex(40, -40, -40, 40, 40, 40)
  endShape()
}

Before:

After:

sketch:

function setup() {
  createCanvas(400, 400, WEBGL);
  background(255)
  noFill()
  stroke(0)
  strokeWeight(20)
  translate(-width/2, -height/2)
  beginShape()
  vertex(20, 20)
  quadraticVertex(
    280, 200,
    100, 380
  )
  endShape()
}

output:

PR Checklist

• npm run lint passes
• Inline documentation is included / updated
• Unit tests are included / updated

[diyaayay]

I think there is redundancy in storing values that is why the tests fail. I will look into it as soon as possible.

[davepagurek]

It looks like the strategy here is to immediately add a cap at the beginning, but I think that might break some other cases, if at a later segment of the stroke, a line connects from vertex 0 to somewhere else, e.g. in a triangle like this:

0      1
x----x
 \  /
  \/
  x
  2

With edges [[0,1], [1,2], [2,0]], I think we'd immediately mark the first vertex as needing a cap, when by the time we get to the end of the edges, we'll see that it's connected again.

I think the problem with curves may be something a bit simpler: when using the debugger on the sketch in the issue, I noticed this for i==0:

I think this means that when we add a curve after an initial vertex, we're outputting two identical vertices at the beginning. Because they're right on top of each other, we can't figure out the direction, so we don't add a cap. Maybe that's because when adding points along a quadratic vertex segment, we start at length 0 along the curve, which is equal to the previous vertex:

p5.js/src/webgl/3d_primitives.js

Line 1910 in 37d3324

let start = 0;

So maybe we just need to start at step instead of 0? (Maybe also for bezierVertex? I haven't checked if it has the same problem as quadratic.)