Skip to content
Switch branches/tags
Go to file
Cannot retrieve contributors at this time
196 lines (155 sloc) 7.11 KB
Author = "Daniel C. Jones"


Gadfly supports advanced plot composition techniques like faceting, stacking, and layering.


It is easy to make multiple plots that all share a common dataset and axis.

using Gadfly, RDatasets
set_default_plot_size(14cm, 8cm) # hide
iris = dataset("datasets", "iris")
plot(iris, xgroup="Species", x="SepalLength", y="SepalWidth",

Geom.subplot_grid can similarly arrange plots vertically, or even in a 2D grid if there are two shared axes.

Subplots have some inner and outer elements, including Guides and Scales. For example, place the guide inside Geom.subplot_grid(...) to change the subplot labels, or outside to change the outer plot labels.

haireye = dataset("datasets", "HairEyeColor")
palette = ["brown", "blue", "tan", "green"]

plot(haireye, y=:Sex, x=:Freq, color=:Eye, ygroup=:Hair,
    Geom.subplot_grid(, orientation=:horizontal),
        Guide.ylabel(orientation=:vertical) ),
    Guide.ylabel("Hair color"), Guide.xlabel("Frequency") )

More examples can be found in the plot gallery at [Geom.subplot_grid](@ref Gallery_Geom.subplot_grid) and [Scale.{x,y}group](@ref Gallery_Scale.xygroup).


To composite plots derived from different datasets, or the same data but different axes, a declarative interface is used. The [Tutorial](@ref Rendering) showed how such disparate plots can be horizontally arranged with hstack. Here we illustrate how to vertically stack them with vstack or arrange them in a grid with gridstack. These commands allow more customization in regards to tick marks, axis labeling, and other plot details than is available with Geom.subplot_grid.

using Gadfly, RDatasets, Compose
iris = dataset("datasets", "iris")
set_default_plot_size(14cm, 16cm) # hide
theme1 = Theme(key_position=:none)
fig1a = plot(iris, x=:SepalLength, y=:SepalWidth, color=:Species, theme1,
          alpha=[0.6], size=:PetalLength, Scale.size_area(maxvalue=7))
fig1b = plot(iris, x=:SepalLength, color=:Species, Geom.density,
          Guide.ylabel("density"), Coord.cartesian(xmin=4, xmax=8), theme1)

hstack and vstack can be composed to create arbitrary arrangements of panels.


If all rows or columns have the same number of panels, it's easiest to use gridstack.

gridstack([p1 p2; p3 p4])

For each of these commands, you can leave a panel blank by passing an empty plot(). Other elements, e.g. Scales and Guides, can be added to blank plots. If the plot contains aesthetic mappings, use Geom.blank.

using Compose # for w, h relative units
set_default_plot_size(21cm, 16cm) # hide
fig1c = plot(iris, x=:SepalWidth, color=:Species, Geom.density,
          Guide.ylabel("density"), Coord.cartesian(xmin=2, xmax=4.5), theme1)
fig1d = plot(iris, color=:Species, size=:PetalLength, Geom.blank,
          Scale.size_area(maxvalue=7), Theme(key_swatch_color="silver"),
          Guide.colorkey(title="Species", pos=[0.55w,-0.15h]),
          Guide.sizekey(title="PetalLength (cm)", pos=[0.2w, -0.10h]))
gridstack([fig1a fig1c; fig1b fig1d])

Note in this example, the Guide pos argument is in [width, height] relative units, which come from Compose.

Lastly, title can be used to add a descriptive string to the top of a stack.

title(hstack(p1,p2), "My creative title")


Introduction: Draw multiple layers onto the same plot by inputing Layer objects to plot.

using Gadfly, RDatasets, Distributions
set_default_plot_size(14cm, 8cm)
iris = dataset("datasets", "iris")
xdata = sort(iris.SepalWidth)
ydata = cumsum(xdata)
line = layer(x=xdata, y=ydata, Geom.line, color=[colorant"red"], 
bars = layer(iris, x=:SepalWidth,
plot(line, bars)

Note that here we used both the DataFrame and AbstractArrays interface to layer, as well a Theme object. See Themes for more information on the latter.

You can share the same DataFrame across different layers:

     layer(x=:SepalLength, y=:SepalWidth),
     layer(x=:PetalLength, y=:PetalWidth, color=[colorant"red"]))

In this case, Gadfly labels the axes with the column names of first layer listed. If this is not what is desired, Guides may be explicitly added.

     layer(x=:SepalLength, y=:SepalWidth),
     layer(x=:PetalLength, y=:PetalWidth, color=[colorant"red"]),
     Guide.xlabel("length"), Guide.ylabel("width"), Guide.title("Iris data"),

Layer inputs: layer() can input Geometries, Statistics, and Themes, but not Scales, Coordinates, or Guides.

There are two rules about layers and Statistics:

  1. Within a layer, all Geoms will use the layer Stat (if it's specified) e.g. layer(Stat.smooth(method=:lm), Geom.line, Geom.ribbon)

  2. For Geoms outside of layers, Gadfly creates a new layer for each Geom, and each Stat is added to the newest layer e.g.

     xdata = range(-9, 9, length=30)
     plot(x=xdata, y=rand(30), Geom.point, Stat.binmean(n=5),
      Geom.line, Stat.step)

Layers and Aesthetics: Aesthetics can also be shared across layers:

plot(iris, Guide.colorkey(title=""),
    layer(x->0.4x-0.3, 0, 8, color=["Petal"]),
    layer(x=:SepalLength, y=:SepalWidth, color=["Sepal"]),
    layer(x=:PetalLength, y=:PetalWidth, color=["Petal"]),
    layer(x=[2.0], y=[4], shape=[Shape.star1], color=[colorant"red"], size=[8pt]),

And layers can inherit aesthetics from the plot:

set_default_plot_size(21cm, 8cm)
p1 = plot(iris, x=:SepalLength, y=:PetalLength,
    layer(Geom.smooth(method=:loess), color=["Smooth"]),
    layer(Geom.point, color=["Points"]))

p2 = plot(iris, x=:SepalLength, y=:PetalLength, color=:Species,
    Geom.smooth(method=:lm), Geom.point, alpha=[0.6],
    layer(Geom.smooth(method=:loess), color=[colorant"grey"], order=2))
hstack(p1, p2)

Note in some layers, it may be better to use specific Geoms e.g. Geom.yerrorbar rather than Geom.errorbar, since the latter will attempt to inherit aesthetics for both Geom.xerrorbar and Geom.yerrobar.

Layer order: the sequence in which layers are drawn, whether they overlap or not, can be controlled with the order keyword. Layers with lower order numbers are rendered first. If not specified, the default order for a layer is 0. Layers which have the same order number are drawn in the reverse order in which they appear in plot's input arguments.

bars = layer(iris, x=:SepalWidth,
line = layer(iris, x=xdata, y=ydata, Geom.line, color=[colorant"red"],
plot(bars, line)