Force colorbuffer to use the size of the screen, not the scene

[asinghvi17]

Description

This also changes the VideoStream constructor similarly, allowing the user to use patterns like this:

CairoMakie.activate!(px_per_unit=3)

record(fig, "streamplot_anim_keener.mp4") do io
    ...
end

This works with GLMakie and CairoMakie, but will probably break WGLMakie since there seems to be no GeometryBasics.width definitions for either WGLDisplay or ThreeDisplay, and I'm not sure how to go about implementing those since the display structs don't contain their Scenes directly.

I built this off my other PR which accelerates recording speed in the presence of rasterized plots, but will rebase once that is merged.

Type of change

Delete options that do not apply:

• [X] New feature (non-breaking change which adds functionality)

Checklist

• [x] Added an entry in NEWS.md (for new features and breaking changes)
• [ ] Added or changed relevant sections in the documentation

[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	20.13s (19.93, 20.37) 0.16+-	22.14s (21.56, 22.82) 0.54+-	20.38s (19.75, 21.02) 0.51+-	19.19ms (18.21, 19.73) 0.52+-	117.74ms (112.65, 119.96) 2.67+-
master	19.99s (19.49, 20.34) 0.37+-	22.21s (21.60, 22.64) 0.39+-	20.24s (19.78, 20.71) 0.37+-	19.36ms (18.88, 19.83) 0.35+-	119.11ms (116.92, 121.57) 1.57+-
evaluation	+0.71%, 0.14s invariant (0.51d, 0.37p, 0.26std)	-0.35%, -0.08s invariant (-0.17d, 0.76p, 0.46std)	+0.72%, 0.15s invariant (0.33d, 0.55p, 0.44std)	-0.90%, -0.17ms invariant (-0.39d, 0.48p, 0.44std)	-1.17%, -1.38ms invariant (-0.63d, 0.27p, 2.12std)
CairoMakie	12.65s (12.48, 12.92) 0.15+-	19.35s (18.80, 20.05) 0.39+-	3.15s (3.08, 3.23) 0.06+-	15.84ms (15.47, 16.22) 0.26+-	26.39ms (25.70, 29.51) 1.38+-
master	12.73s (12.54, 12.91) 0.12+-	19.34s (18.89, 19.66) 0.26+-	3.10s (3.03, 3.16) 0.04+-	15.75ms (15.33, 16.08) 0.26+-	25.97ms (25.73, 26.64) 0.31+-
evaluation	-0.69%, -0.09s invariant (-0.64d, 0.25p, 0.14std)	+0.04%, 0.01s invariant (0.02d, 0.96p, 0.32std)	+1.66%, 0.05s invariant (1.06d, 0.07p, 0.05std)	+0.58%, 0.09ms invariant (0.36d, 0.52p, 0.26std)	+1.59%, 0.42ms invariant (0.42d, 0.46p, 0.85std)
WGLMakie	20.65s (19.81, 22.64) 1.05+-	27.98s (26.86, 30.38) 1.51+-	45.55s (43.69, 49.51) 2.28+-	23.45ms (22.45, 24.87) 0.98+-	1.77s (1.71, 1.83) 0.05+-
master	20.76s (20.08, 22.30) 0.95+-	27.83s (27.05, 29.30) 0.99+-	45.28s (43.82, 48.22) 1.89+-	24.10ms (22.14, 25.98) 1.55+-	1.78s (1.70, 1.83) 0.04+-
evaluation	-0.49%, -0.1s invariant (-0.10d, 0.85p, 1.00std)	+0.53%, 0.15s invariant (0.12d, 0.83p, 1.25std)	+0.59%, 0.27s invariant (0.13d, 0.81p, 2.08std)	-2.74%, -0.64ms invariant (-0.50d, 0.38p, 1.27std)	-0.58%, -0.01s invariant (-0.22d, 0.68p, 0.05std)

[jkrumbiegel]

I don't really understand where in the code the px_per_unit attribute is picked up by colorbuffer, is it possible to pass it to record as well, just like you can do it for save? That's what I'd want, I mainly see the hidden state as a convenience while plotting, but it should always be directly override-able.

[asinghvi17]

Hmm, that's a good idea - might require a bit of refactor but will do that.

The px_per_unit is in this case read from the backend in backend_show, and in order to pass it on I'll refactor backend_display to pass some kwargs on to backend_show through an IOContext.

[SimonDanisch]

closed in favor of #2306