CSS Color 4 color spaces for canvasGradient

[mysteryDate]

Add ColorInterpolationMethod, HueInterpolationMethod and premultiplyAlpha attributes to CanvasGradient.

• At least two implementers are interested (and none opposed):

  • Chromium
  • Webkit

• Tests are written and can be reviewed and commented upon.

• Implementation bugs are filed:

  • Chromium
  • Gecko: …
  • WebKit: …

• Corresponding HTML AAM & ARIA in HTML issues & PRs:

• MDN issue is filed: …

• The top of this comment includes a clear commit message to use.

Open questions:

• Should we add an "options" object at the end, to specify these attributes at creation? (Method 2 in Define color space for interpolation interface for canvas gradients #8296)
• Do we expose color stops, while we're at it?

TODO:

• Lint
• Mozilla/Apple stances, bugs
• Experimental implementation in Chromium
• WPT tests

---

/canvas.html ( diff )

[annevk]

Can you please restore the PR template?

cc @whatwg/canvas

[annevk]

This lacks normative definitions for the new attributes.

[yisibl]

https://crbug.com/394052224 This cannot be opened by non-Google internal users.

[mysteryDate]

https://crbug.com/394052224 This cannot be opened by non-Google internal users.

Sorry! I believe I have fixed this.

[mysteryDate]

Can you please restore the PR template?

cc @whatwg/canvas

Done!

[mysteryDate]

This lacks normative definitions for the new attributes.

Just added some basic ones. Do we need to define the color spaces themselves? premultipliedAlpha is defined elsewhere in this document, but color interpolation spaces are only defined pretty tersely:

https://html.spec.whatwg.org/#colour-spaces-and-colour-correction

And hue interpolation does appear to be defined at all.

Should my normative definitions cite CSS? Or do we need separate definitions here.

[annevk]

1. These attributes need to be defined in a way that is similar to how fillStyle is defined. I.e., some internal concept to hold the data, getter steps, and setter steps.
2. The internal concept needs to be referenced from where we need to use it for some calculation (i.e., from the processing model, to the extent one is defined).

I think it's okay if some aspects are a bit hand-wavy, but ideally they are not or at least done in such a way that we can gradually fill out the details over time.

If concepts are shared with CSS it would be good to reference CSS. Perhaps some kind of shared abstraction is in order. We have done this for other features that are shared with CSS as well.