Merge branch 'master' into ff/lighting

docs/explanations/recipes.md

@@ -47,7 +47,10 @@ Defining a conversion for every plot type likely won't make sense, so one can re
 Makie.convert_arguments(P::Type{<:Scatter}, ::MyType) = convert_arguments(P, rand(10))
 ```
 
-Conversion traits make it easier to define behavior for a group of plot types that share the same trait. `PointBased` for example applies to `Scatter`, `Lines`, etc. Predefined are `NoConversion`, `PointBased`, `SurfaceLike` and `VolumeLike`.
+Conversion traits make it easier to define behavior for a group of plot types that share the same trait.
+`PointBased` for example applies to `Scatter`, `Lines`, etc.
+The predefined traits are `NoConversion`, `PointBased`, `CellGrid <: GridBased`, `VertexGrid <: GridBased`, `ImageLike`, `VolumeLike` and `SampleBased`.
+They all inherit from `ConversionTrait`, sometimes indirectly.
 
 ```julia
 Makie.convert_arguments(::PointBased, ::MyType) = ...

src/basic_recipes/streamplot.jl

@@ -28,7 +28,7 @@ See the function `Makie.streamplot_impl` for implementation details.
         maxsteps = 500,
         color = norm,
 
-        arrow_size = 15,
+        arrow_size = automatic,
         arrow_head = automatic,
         density = 1.0,
         quality = 16,
@@ -213,10 +213,22 @@ function plot!(p::StreamPlot)
         rotations = map(x -> x[2], data)
     end
 
+    arrow_size = map(p, p.arrow_size) do arrow_size
+        if arrow_size === automatic
+            if N == 3
+                return 0.2 * minimum(p.limits[].widths) / minimum(p.gridsize[])
+            else
+                return 15
+            end
+        else
+            return arrow_size
+        end
+    end
+
     scatterfun(N)(
         p,
         lift(first, p, data);
-        markersize=p.arrow_size, rotations=rotations,
+        markersize=arrow_size, rotations=rotations,
         color=lift(x -> x[4], p, data),
         marker = lift((ah, q) -> arrow_head(N, ah, q), p, p.arrow_head, p.quality),
         colormap_args...,