Fix bezier marker stroke

[ffreyer]

Description

Without the ClosePath stroke outlines are incomplete in CairoMakie.

Type of change

• Bug fix (non-breaking change which fixes an issue)

Checklist

• Added an entry in NEWS.md (for new features and breaking changes)
• Added or changed relevant sections in the documentation
• Added unit tests for new algorithms, conversion methods, etc.
• Added reference image tests for new plotting functions, recipes, visual options, etc.

[MakieBot]

Compile Times benchmark

Note, that these numbers may fluctuate on the CI servers, so take them with a grain of salt. All benchmark results are based on the mean time and negative percent mean faster than the base branch. Note, that GLMakie + WGLMakie run on an emulated GPU, so the runtime benchmark is much slower. Results are from running:

using_time = @ctime using Backend
# Compile time
create_time = @ctime fig = scatter(1:4; color=1:4, colormap=:turbo, markersize=20, visible=true)
display_time = @ctime Makie.colorbuffer(display(fig))
# Runtime
create_time = @benchmark fig = scatter(1:4; color=1:4, colormap=:turbo, markersize=20, visible=true)
display_time = @benchmark Makie.colorbuffer(display(fig))

	using	create	display	create	display
GLMakie	13.46s (13.31, 13.64) 0.11+-	1.31s (1.28, 1.34) 0.02+-	939.27ms (915.48, 987.84) 23.58+-	13.45ms (13.33, 13.55) 0.08+-	104.60ms (103.27, 106.23) 1.10+-
master	13.45s (13.27, 13.58) 0.12+-	1.32s (1.27, 1.36) 0.03+-	940.31ms (889.57, 959.17) 25.13+-	13.44ms (13.18, 13.65) 0.15+-	105.09ms (102.48, 108.04) 2.01+-
evaluation	+0.12%, 0.02s invariant (0.14d, 0.80p, 0.11std)	-0.54%, -0.01s invariant (-0.26d, 0.64p, 0.03std)	-0.11%, -1.04ms invariant (-0.04d, 0.94p, 24.36std)	+0.04%, 0.0ms invariant (0.04d, 0.94p, 0.12std)	-0.46%, -0.48ms invariant (-0.30d, 0.59p, 1.56std)
CairoMakie	13.75s (13.56, 14.16) 0.21+-	1.64s (1.60, 1.68) 0.03+-	351.29ms (340.52, 373.00) 12.27+-	16.16ms (15.98, 16.36) 0.13+-	8.56ms (8.40, 8.81) 0.13+-
master	14.05s (13.78, 14.65) 0.29+-	1.64s (1.56, 1.74) 0.06+-	400.83ms (375.99, 465.56) 31.06+-	16.46ms (16.09, 16.96) 0.34+-	8.59ms (8.39, 8.71) 0.12+-
evaluation	-2.19%, -0.3s faster ✓ (-1.19d, 0.05p, 0.25std)	-0.11%, -0.0s invariant (-0.04d, 0.95p, 0.05std)	-14.10%, -49.54ms faster✅ (-2.10d, 0.00p, 21.67std)	-1.86%, -0.3ms invariant (-1.18d, 0.06p, 0.23std)	-0.37%, -0.03ms invariant (-0.25d, 0.65p, 0.13std)
WGLMakie	14.10s (13.85, 14.31) 0.14+-	1.39s (1.32, 1.46) 0.05+-	12.74s (12.56, 12.91) 0.15+-	16.87ms (15.39, 22.50) 2.53+-	946.35ms (902.22, 1076.44) 59.46+-
master	14.25s (14.00, 14.45) 0.16+-	1.44s (1.40, 1.48) 0.03+-	12.75s (12.51, 13.08) 0.18+-	16.08ms (14.41, 17.75) 1.32+-	916.47ms (874.04, 936.49) 21.68+-
evaluation	-1.02%, -0.14s invariant (-0.96d, 0.10p, 0.15std)	-3.69%, -0.05s faster ✓ (-1.27d, 0.04p, 0.04std)	-0.05%, -0.01s invariant (-0.04d, 0.94p, 0.17std)	+4.63%, 0.78ms invariant (0.39d, 0.49p, 1.93std)	+3.16%, 29.88ms invariant (0.67d, 0.25p, 40.57std)

[ffreyer]

I added another refimg that includes stroke. It seems like the signed distance field derived for GLMakie is also kind of messed up. Looks like the what's happening with 2D scaling in #2950... Are we introducing that along the way? (Also the sizes differ)

CairoMakie:

GLMakie:

[jkrumbiegel]

Huh didn't the sizes match at one point in the past?

[ffreyer]

I don't think we checked with stroke. I think I've seen that with other stroked markers though. Or maybe lines.

On the bad stroke in GLMakie - this is happening because the sdf is generated with a different aspect ratio here

Makie.jl/src/bezier.jl

Line 500 in 08d80e9

path_unit_square = fit_to_unit_square(path_replaced, false)

which results in different width and height scaling here

Makie.jl/src/utilities/texture_atlas.jl

Line 482 in 08d80e9

path_width = widths(Makie.bbox(path))

[ffreyer]

This will need a new texture atlas. Should I just increment this?

Makie.jl/src/utilities/texture_atlas.jl

Line 73 in 08d80e9

const SERIALIZATION_FORMAT_VERSION = "v1"

[github-actions]

Missing reference images

Found 1 new images without existing references.
Upload new reference images before merging this PR.

[ffreyer]

With e86aee6 (Circle scatter to mark center and debug boxes turned on in distance_shape.frag)

With 245fea7:

[github-actions]

Missing reference images

Found 1 new images without existing references.
Upload new reference images before merging this PR.

[jkrumbiegel]

Should I see a difference between them? :)

[ffreyer]

GLMakie master: vline has blurry edge and hline has shrunken stroke/glow
patch 1: fix scaling of stroke/glow/AA area
patch 2: reduce size of quad (gray background in image, also size is texture atlas)

[jkrumbiegel]

ahh now I see it, nice!

[SimonDanisch]

This will need a new texture atlas. Should I just increment this?

Yes, I think that should be enough!

[github-actions]

Missing reference images

Found 1 new images without existing references.
Upload new reference images before merging this PR.

[github-actions]

Missing reference images

Found 1 new images without existing references.
Upload new reference images before merging this PR.

[github-actions]

Missing reference images

Found 1 new images without existing references.
Upload new reference images before merging this PR.

[github-actions]

Missing reference images

Found 1 new images without existing references.
Upload new reference images before merging this PR.

[github-actions]

Missing reference images

Found 1 new images without existing references.
Upload new reference images before merging this PR.