method <SmoothGraphics.bezierCurveTo> error

[pixijs]

method error

from graphics-smooth.

[dev7355608]

Please share some code with me that allows me to reproduce this error.

[ivanpopelyshev]

OMG how! please demo

[dev7355608]

@ivanpopelyshev Today a non-polygon hole in a polygon shape was giving me the same error. I tracked that down to the fact that the holes use the line command of the shape instead of the one matching its type. Fixed that. But then I noticed why it failed so terribly: because data.closeStroke was false. Then I tried to draw non-closed polygons: those are not rendering at all. Something goes wrong in PolyBuilder#line if closeStroke is false.

let p = new PIXI.Polygon([0, 0, 100, 0, 100, 100, 0, 100]);

p.closeStroke = false;

let g = app.stage.addChild(new PIXI.smooth.SmoothGraphics()
        .lineStyle(1)
        .beginFill(0xffffff)
        .drawPolygon(p)
        .endFill());

[ivanpopelyshev]

Oh, interesting

[dev7355608]

@ivanpopelyshev I figured out why this error happens and fixed it with a19cda7: joint should be endJoint. The code produced a FILL joint, because it increased a NONE joint by one if closeStroke was false; therefore hasTriangle became true in the segment packer, which corrupted index buffer.

But after changing joint to endJoint there is still something wrong with the MITER joint:

I'm not quite sure what exactly is happening here in this part of the code. If I comment it out, the line is rendered correctly.

graphics-smooth/src/shapes/PolyBuilder.ts

Lines 284 to 297 in a19cda7

	if (endJoint === JOINT_TYPE.JOINT_MITER)
	{
	let jointAdd = 0;
	if (dx3 * dx + dy3 * dy > -eps)
	{
	jointAdd++;
	}
	if (endJoint === JOINT_TYPE.JOINT_MITER && dx2 * dx + dy2 * dy > -eps)
	{
	jointAdd += 2;
	}
	endJoint += jointAdd;
	}

[dev7355608]

@ivanpopelyshev I'm considering to remove this code entirely:

graphics-smooth/src/shapes/PolyBuilder.ts

Lines 260 to 305 in 994fb4d

	if (joint >= JOINT_TYPE.JOINT_BEVEL && joint <= JOINT_TYPE.JOINT_MITER)
	{
	const dx2 = nextX - x2;
	const dy2 = nextY - y2;
	if (endJoint >= JOINT_TYPE.JOINT_BEVEL
	&& endJoint <= JOINT_TYPE.JOINT_MITER + 3)
	{
	const D = dx2 * dy - dy2 * dx;
	if (Math.abs(D) < eps)
	{
	switch (joint & ~3)
	{
	case JOINT_TYPE.JOINT_ROUND:
	endJoint = JOINT_TYPE.JOINT_CAP_ROUND;
	break;
	default:
	endJoint = JOINT_TYPE.JOINT_CAP_BUTT;
	break;
	}
	}
	}
	if (joint === JOINT_TYPE.JOINT_MITER)
	{
	let jointAdd = 0;
	if (dx3 * dx + dy3 * dy > -eps)
	{
	jointAdd++;
	}
	if (endJoint === JOINT_TYPE.JOINT_MITER && dx2 * dx + dy2 * dy > -eps)
	{
	jointAdd += 2;
	}
	endJoint += jointAdd;
	}
	}
	if (prevCap === 0)
	{
	if (Math.abs(dx3 * dy - dy3 * dx) < eps)
	{
	prevCap = JOINT_TYPE.CAP_BUTT2;
	}
	}

It seems to work fine without it, and it is flawed, but I'm not so sure exactly what it is supposed to do. Do you think it's safe to delete this code?

[ivanpopelyshev]

Oh, hi! Thanks for checking up on me.

Yes, its one of things i coded in beginning that i didnt fix yet. You can remove it if you test degenerate cases like "line goes there and back"

UPD: i think it was idea like "if bevel is good enough, use good miter instead, without a joint". It shouldnt affect things because we already are using instancing, right? or did i forget to enable it here?

[dev7355608]

The degenerate test in the examples looks good if I remove this part of the code. I tried all join and cap combinations. Then I can proceed to remove it?

I'm not very familiar with this joint stuff yet. MITER + 3 is higher quality? But this looked odd to me:

verts[JOINT_TYPE.JOINT_MITER] = 4 + 5;
verts[JOINT_TYPE.JOINT_MITER + 1] = 4 + 5;
verts[JOINT_TYPE.JOINT_MITER + 2] = 4;
verts[JOINT_TYPE.JOINT_MITER + 3] = 4;

+2 and +3 have less vertices. This the only joint type that has different vertex counts.

[ivanpopelyshev]

Yes, it was there to save vertices. I agree to remove it because , whether instancing is implemented or not - it will make this "optimization" worseless soon

[dev7355608]

Fixed in 0.0.26.