Extend color for tuple of RGB components

[t-bltg]

Allow passing arbitrary colors, e.g.:

using UnicodePlots

lineplot([1, 2], color=:black)
lineplot([1, 2], color=(0, 100, 200))
lineplot([1, 2], color=255)

Needed for Plots.jl, since we use RGB values internally.
Uses Crayon for color conversion: I think this is appropriate, since Crayon is already a UnicodePlots dependency used in Heatmaps, and its ANSIColor type is coherent with terminal color output.

With #136 applied, all tests pass with newly generated images:

Test Summary: | Pass  Total
tst_common.jl |  106    106
Test Summary:   | Pass  Total
tst_graphics.jl |   79     79
Test Summary: | Pass  Total
tst_canvas.jl |  140    140
Test Summary: | Pass  Total
tst_plot.jl   |   64     64
Test Summary:  | Pass  Total
tst_barplot.jl |   32     32
Test Summary:    | Pass  Total
tst_histogram.jl |   19     19
Test Summary:      | Pass  Total
tst_scatterplot.jl |   45     45
Test Summary:   | Pass  Total
tst_lineplot.jl |   79     79
Test Summary: | Pass  Total
tst_spy.jl    |   16     16
Test Summary:  | Pass  Total
tst_boxplot.jl |   20     20
Test Summary:  | Pass  Total
tst_heatmap.jl |   75     75

Fix #20.
Fix #58.

[t-bltg]

@johnnychen94, this seems a bit hacky, should I PR to Crayons.jl instead to have stable API conversions to 256 color ?

[johnnychen94]

Proposing a stable API in Crayons sounds okay to me but I know very little about this. I don't know if it's useful, there are some discussions on JuliaGraphics/Colors.jl#473

[t-bltg]

Thanks for the link. What we need in UnicodePlots is something generic working on all terminals. Since we use printstyled(...) I think the implementation in this PR is ok using 0-255 integers.

@KristofferC, do you have any advice for this ?

[t-bltg]

The CI failures are related to base/util.jl inserting 38;5; to set the foreground color, when using integers in the range 0-255 even if visually the images are correct. I must thus regenerate reference images.

Except for Heatmap, for which this PR fixes a regression, dotcanvas & heatmapcanvas are now similar in terms of colors for the same data:

[johnnychen94]

Sorry for the back and forth. I just noticed that this PR doesn't have associated test cases added. Can you also add some tests for the new keyword types?

Currently, when I try barplot(["B", "A"], [2, 1]; color=(255, 0, 0)) it's broken. Looks like something is missing here?

[t-bltg]

I just noticed that this PR doesn't have associated test cases added

Good point, will do.

Can you also add some tests for the new keyword types?

I don't understand, could you precise which keyword types you're talking about ?

Currently, when I try barplot(["B", "A"], [2, 1]; color=(255, 0, 0)) it's broken.

Good catch ! Will fix.

[johnnychen94]

Can you also add some tests for the new keyword types?

I don't understand, could you precise which keyword types you're talking about ?

Sorry for the inaccuracy, I meant to test NTuple{3, Integer} types for color keyword.

[t-bltg]

Sorry for the inaccuracy, I meant to test NTuple{3, Integer} types for color keyword.

Oh ok, yes 👍

[t-bltg]

I've inserted some colors tests covering NTuple{3, Integer} and Integer in tst_canvas, tst_barplot, tst_boxplot, tst_histogram.

We now have 3 colors type: UserColorType (broad), ColorType for UnicodePlots canvas, and finally JuliaColorType for printstyled.

After visual inspection against master using:

root="$HOME/Downloads/UnicodePlots.jl"  # fork

for f in $(find $root/test/references -name '*.txt'); do
 echo; echo
 echo "== local($f) =="
 cat $f
 echo
 url="https://raw.githubusercontent.com/Evizero/UnicodePlots.jl/master/${f#$root/}"
 echo "== remote($url) =="
 wget -q -O - $url
done

, I think this PR can be reviewed again.

[johnnychen94]

Sorry, I didn't look very carefully but the test looks good to me. Since a lot of references are changed so I assume there will be some implicit behavior differences, I'd like to bump a minor version to indicate some "breaking" changes, does it sound good to you? @t-bltg

[t-bltg]

Sorry, I didn't look very carefully but the test looks good to me. Since a lot of references are changed so I assume there will be some implicit behavior differences, I'd like to bump a minor version to indicate some "breaking" changes, does it sound good to you? @t-bltg

Yes, I'm good with the minor bump. I will apply the changes in Plots.jl accordingly after the release.

[johnnychen94]

It's actually a major version bump 😄 v2.0.0