Linestyle aes

[Mattriks]

• I've added an entry to NEWS.md
• I've updated the documentation to reflect these changes
• I've added and/or updated the unit tests
• I've run the regression tests
• I've squash'ed or fixup'ed junk commits with git-rebase
• I've built the docs and confirmed these changes don't cause new errors

Linestyle aes

This PR adds:

1. linestyle as an aesthetic (Feature request aes(linetype=variable)? #455), not just as a Theme field (Dashed Lines #392)
2. Scale.linestyle_discrete, so 1. & 2. are steps towards Guide.linekey (Legend with Shapes - Colorblind Accessibility #1042)
3. A new render(geom::LineGeometry) function which fixes Geom.line not drawn after NaNs in SVGs #903, and handles multiple aesthetics smoothly.
4. Also closes BoundsError with group aesthetic #967, closes BoundsError with plot(fs::Array, a,b, Scale.y_log10) #755

Note that I'll do Guide.linekey in a separate PR.

Examples:

labs = ["exp", "sqrt", "log", "winsor", "linear"]
funcs=[x->60*(1-exp.(-0.2*x)), x->sqrt.(x)*10, x->log.(x)*10, x->winsor(x, prop=0.15), x->x*0.6]
colors = ["α","α","β","γ","γ"] 
x = [1.0:30;]
D = vcat([DataFrame(x=x, y=f(x), linev=l, col=c) for (f,l,c) in zip(funcs, labs, colors)]...)
D[134:136,:y] = NaN

pa = plot(D, x=:x, y=:y, linestyle=:linev, Geom.line, Guide.title("linestyle=:linev"))
# Two aesthetics:
pb = plot(D, x=:x, y=:y, linestyle=:linev, color=:col, Geom.line,
   Scale.linestyle_discrete(levels=["exp", "log", "sqrt", "winsor", "linear"]),
    Guide.title("linestyle≠color"))
# Multiple aesthetics:
pc = plot(dataset("datasets", "CO2"), x=:Conc, y=:Uptake, 
    group=:Plant, linestyle=:Treatment, color=:Type, Geom.line,
    Scale.linestyle_discrete(order=[2,1]),
    Theme(key_position=:top, key_title_font_size=-5mm) )

hstack(pa,pb,pc)

[tlnagy]

This is awesome. I think the biggest thing would be to have the corresponding legend info like in #1156. Would you prefer to have that in a separate PR?

[bjarthur]

looks great! the only thing i notice is that all the other gallery examples use 21cm for multi-panel figures and 14cm for single.

[Mattriks]

I'll probably expand it to 21cm after I implement Guide.linekey, since a linekey will appear for both plots. For now its 18cm to signify future growth!

[Mattriks]

Yes I'll do Guide.linekey in a separate PR (which I did mention in my original post).

[codecov-io]

Codecov Report

Merging #1181 into master will decrease coverage by 0.15%.
The diff coverage is 91.66%.

@@            Coverage Diff             @@
##           master    #1181      +/-   ##
==========================================
- Coverage   87.33%   87.17%   -0.16%     
==========================================
  Files          34       34              
  Lines        4081     4048      -33     
==========================================
- Hits         3564     3529      -35     
- Misses        517      519       +2

Impacted Files	Coverage Δ
src/Gadfly.jl	79.04% <ø> (ø)	⬆️
src/data.jl	0% <ø> (ø)	⬆️
src/aesthetics.jl	82.35% <ø> (ø)	⬆️
src/geom/line.jl	98.07% <100%> (+7.85%)	⬆️
src/geom/hvabline.jl	94.11% <100%> (ø)	⬆️
src/scale.jl	88.26% <100%> (+0.05%)	⬆️
src/geom/polygon.jl	66.66% <100%> (ø)	⬆️
src/geom/segment.jl	94.59% <100%> (ø)	⬆️
src/theme.jl	66.1% <64.28%> (-1.21%)	⬇️
src/misc.jl	51.87% <0%> (-3.75%)	⬇️
... and 2 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 338fe77...180a833. Read the comment docs.

[Mattriks]

I've also tested the strange case in #967 with this PR, and it works:

# baddf.csv from #967
baddf = CSV.read("baddf.csv", allowmissing=:none)
pa = plot(baddf, x=:phi, y=:fhat, color=:i, Geom.line, Scale.color_discrete)
pb =  plot(baddf, x=:phi, y=:fhat, group=:i, Geom.line)
hstack(pa,pb)

So this PR is ready to merge and will also close #967.

[Mattriks]

Also closes #755:

p2 = plot([sin,cos],0,1, Scale.y_log10, Theme(key_position=:inside)) 

[bjarthur]

thanks for this! looks great. i'm going to go ahead and merge, but in future could you please squash your junk commits like the second one here? they just add noise when reading through the commit history, and make things more difficult to revert, cherry-pick, or bisect.

[tlnagy]

@Mattriks have you had a chance to make the Guide.linekey PR?

[Mattriks]

Not yet, that should go on the next roadmap, after #1385 (which I'm hoping to get back to soon)