#35 Implement miter_limit_join

[o0Ignition0o]

This is Working on a function signature, before miter_limit_join_implementation (and tests).
I guess I should remove the magic number 4, to use an enum or something instead.

This should handle #35 some day x)

[o0Ignition0o]

Hey @nical , feel free to tell me if:

• There's a better way to represent the svg specification than using consts
• Panicking during the obj construction is fine by you, or if i should set default value instead ( I don't really understand if it's the StrokeBuilder's duty to handle exceptions as described in https://svgwg.org/svg2-draft/implnote.html#ErrorProcessing )
• I m allowed to link to the official svg spec in the comments ?

Thaks a lot :D

[o0Ignition0o]

else if join_type == LineJoin::Miter && normal.square_length() > limit * limit { // Per SVG spec: If the stroke-miterlimit is exceeded, the line join // falls back to bevel. join_type = LineJoin::Bevel } Ok it looks like I m' reimplementing something you already have ^^' I'll refactor my code to use your version of "stroke-miterlimit is exceeded"

[o0Ignition0o]

I guess I'm almost there, still need to handle a weird glitch on the top left corner:
No idea whether it's related to #166 (comment) or not :/
[Edit] : After a couple of tryouts, it looks like I'm responsible for the glitch. Will try to fix it asap :)
[Edit 2] : I'm going a bit crazy about this issue, I think I need help x)
Here's what I get with StrokeOptions::tolerance(0.022).with_line_join(LineJoin::MiterClip).with_miter_limit(2.0)

And here's what I get with StrokeOptions::tolerance(0.022).with_line_join(LineJoin::MiterClip).with_miter_limit(2.0).dont_apply_line_width()

In all screenshots I build the following example path :

pub fn build_test_path<Builder: SvgBuilder>(path: &mut Builder) {; path.move_to(point(0.0, 0.0)); path.relative_line_to(vec2(10.0, 10.0)); path.relative_line_to(vec2(-6.0, -10.0)); path.close(); }
Any thoughts ? :)

BTW I'd love to add some tests to it, but I dont' really know If it's relevant, and how to proceed ^^'

[nical]

nit: let's use block indentation here, like this:

        let v = add_vertex!(
            self,
            Vertex {
                position: self.current,
                normal: normal,
                advancement: self.length,
                side: front_side,
            }
        );

[nical]

nit: really a minor thing but I tend to prefer when the complex code is out of the nested blocks, for example:

if condition {
    // complicated stuff
    return ...
}
return simple_stuff()

could be implemented like:

if !condition {
    return simple_stuff()
}
// complicated stuff
return ...

It helps with keeping the cyclomatic complexity low (what a scary word!) and and in my experience the code looks a tad less intimidating that way.

[nical]

BTW I'd love to add some tests to it, but I dont' really know If it's relevant, and how to proceed ^^'

Yeah, I have put a lot of focus on the fill tessellator so the latter has a pretty great test coverage but it would be good to start thinking about what kind of testing would be useful for the stroke tessellator.
I think that's more of a general thing, to look into outside the scope of this PR.

[o0Ignition0o]

I have to do an other pass checking for unused variables + silly stuff that could be optimized away, but I feel like it should work fine :)

[o0Ignition0o]

There might still be something wrong though, since I don't really understand the "neg_if_right" part of the bevel implementation, so I did not use it (and I might mix up start_vertex and end_vertex i guess?)

[o0Ignition0o]

This should be it ! :)

[nical]

Looks good!

[nical]

Oops actually there was a bug here: this should be front_normal instead of normal, I'll land a quick fix.