🌈 Gradient - Tracking issue

[Dherse]

Gradient progress report:

• Part 1: Color rework Gradient Part 1 - Color rework #2171
• Part 1a: Color fixes Gradient Part 1a - Color fixes #2262
• Part 2: Linear gradients Gradient Part 2 - Linear gradients #2279
• Part 2a: Sharp gradient fixes Gradient Part 2a - Fix sharp gradients in SVG #2307
• Part 3: Radial gradients Gradient Part 3 - Radial gradients #2312
• Part 4: Conic gradients Gradient Part 4 - Conic gradients #2325
• Part 5a: Gradients on text (relative: "parent" aka single gradient on all of the text) Part 5a: Gradients on text with relative: auto or relative: "parent" #2364
• Part 5b: Gradient on text (relative: "self" aka per-glyph gradients)
• Part 5b: Fix gradient on math issue Gradient Part 5b: Fix gradients in math #2604 Gradient Part 5c: Fix gradient rotation on text & math #2606
• Part 6: Pattern fills Gradient Part 6 - Pattern fills #2740
• Part 7: Gradient transparency
• Part 8: Gradient & Pattern AA in rasterized output
• Part 8: Additional gradient features:
  • Hue rotation control
  • Ability to control linear gradient as a line (start and end point instead of angle)
  • Ability to control linear gradient start point
  • Ability to deform radial and conical gradient (allowing for elongated shapes)
• Part 9: Gradients along a path (maybe)

The vast majority of those features (all of them in SVG) are available on my gradient branch. However, in this branch the layout engine is completely broken, which I have fixed in these PRs, but didn't bother to on my experimental branch.

The reason why gradients are split up is because part 1 and 2 already account for close to 5000 lines added, the size and complexity of gradients simply balloons when you consider that we support three different export targets, each with their own quirks and limitations that need to be worked around.

This is just a tracking issue I can point to in other gradient PRs ;)

[Dherse]

Also closes #443 (with Part 8).

[Dherse]

Also closes #260. I'd even go as far as stating that it already closes it.

[Dherse]

It was decided in discussion with @laurmaedje that we would not support (at least now) glyph-by-glyph gradients as they're extremely heavy file-size wise, kind of ugly, and difficult to get right.

Starting here on Discord: https://discord.com/channels/1054443721975922748/1088371867913572452/1161943386048188436

[stellarpower]

This looks fab! Would building it be as simple as pointing rust to a crate from your fork? Afraid I've not much experience with the language. I would like some gradients in my document, so I can do some basic testing.

[laurmaedje]

@stellarpower Most of the gradient work has already landed on main, so if you just compile Typst from main you can just use it.

[stellarpower]

Alright, thanks! I will just try building the master branch then.

[stellarpower]

As of a596663, it seems that I can fill a box with a gradient (fill can take a gradient) - but the stroke for lines doesn't seem to support this; is this worth adding as a subitem above? I imagine single straight lines ought to be a simpler starting point than arbitrary paths or font glyphs, and would enable people to take it for a spin in more situations!

[Dherse]

@stellarpower With regards to line, they generally you work, you can find an example here but I generally recommend keeping simple linear gradients on line as line technically have a zero width which makes other gradient look... funky?

I imagine single straight lines ought to be a simpler starting point than arbitrary paths or font glyphs

Truth be told, there is essentially no difference wrt. the "complexity" of the stroke for gradient fills. Currently, gradient fills in parent mode will work on any path you throw at them, it's only when applied in relative: "self" or relative: "auto" on lines that it can get tricky.

I should also mention that my gradient branch has the entire layout engine broken, I strongly advice not to try using it. It just contains the experimental code I have been working on since June/early July

[stellarpower]

Okay, so I have to do the 1pt + bit, I presume that turns it from a gradient to a stroke.

It seems that fill accepts a gradient or a color, whilst line accepts a color or a stroke (which can be based off a gradient itself), but not a gradient.

Thanks for the advice on your branch, I'll stay clear then.

[emilyyyylime]

Okay, so I have to do the 1pt + bit, I presume that turns it from a gradient to a stroke.

Also consider stroke: (paint: gradient(..))

[laurmaedje]

I think we just need to add the missing gradient -> stroke cast. Seems like an oversight.

[laurmaedje]

Should we close this?

[Dherse]

I was planning on working on some more features, namely pattern fills in the upcoming weeks just haven't found the motivation just yet