WIP: Beeswarm plot

README.md

@@ -14,6 +14,7 @@ This package contains many statistical recipes for concepts and types introduced
     - histogram/histogram2d
     - boxplot
     - violin
+    - beeswarm
     - marginalhist
     - corrplot/cornerplot
 
@@ -25,18 +26,24 @@ using StatPlots
 gr(size=(400,300))
 ```
 
-The `DataFrames` support allows passing `DataFrame` columns as symbols. Operations on DataFrame column can be specified using quoted expressions, e.g.
+The `DataFrames` support allows passing `DataFrame` columns as
+symbols. Operations on `DataFrame` column can be specified using quoted
+expressions, e.g.
+
 ```julia
 using DataFrames
 df = DataFrame(a = 1:10, b = 10*rand(10), c = 10 * rand(10))
 plot(df, :a, [:b :c])
 scatter(df, :a, :b, markersize = :(4 * log(:c + 0.1)))
 ```
+
 If you find an operation not supported by DataFrames, please open an issue. An alternative approach to the `StatPlots` syntax is to use the [DataFramesMeta](https://github.com/JuliaStats/DataFramesMeta.jl) macro `@with`. Symbols not referring to DataFrame columns must be escaped by `^()` e.g.
+
 ```julia
 using DataFramesMeta
 @with(df, plot(:a, [:b :c], colour = ^([:red :blue])))
 ```
+
 ---
 
 ## marginalhist with DataFrames

src/StatPlots.jl

@@ -1,4 +1,3 @@
-
 module StatPlots
 
 using Reexport
@@ -24,12 +23,11 @@ include("cornerplot.jl")
 include("distributions.jl")
 include("boxplot.jl")
 include("violin.jl")
+include("beeswarm.jl")
 include("hist.jl")
 include("marginalhist.jl")
 include("bar.jl")
 include("shadederror.jl")
 include("groupederror.jl")
 
-
-
 end # module

src/beeswarm.jl

@@ -0,0 +1,66 @@
+@shorthands beeswarm
+
+# ---------------------------------------------------------------------------
+# Beeswarm plot
+@recipe function f(::Type{Val{:beeswarm}}, x, y, z; trim=false, side=:both)
+
+  side = check_side(side)
+
+  xp, yp = Float64[], Float64[]
+  glabels = sort(collect(unique(x)))
+  bw = d[:bar_width]
+  bw == nothing && (bw = 0.8)
+
+  for (i,glabel) in enumerate(glabels)
+    # We get the values for this label
+    lab_y = y[filter(i -> _cycle(x,i) == glabel, 1:length(y))]
+    lab_x = zeros(lab_y)
+
+    # Number of bins (defaults to sturges)
+    binning_mode = d[:bins]
+    if binning_mode == :auto
+      binning_mode = :sturges
+    end
+    n = Plots._auto_binning_nbins(tuple(lab_y), 1, mode=binning_mode)
+
+    # Get the widths and the coordinates
+    widths, centers = StatPlots.violin_coords(lab_y, trim=trim, n=n)
+    isempty(widths) && continue
+
+    # normalize
+    hw = 0.5Plots._cycle(bw, i)
+    widths = hw * widths / Plots.ignorenan_maximum(widths)
+
+    # make the violin
+    xcenter = Plots.discrete_value!(d[:subplot][:xaxis], glabel)[1]
+
+    for i in 2:length(centers)
+      inside = Bool[centers[i-1] < u <= centers[i] for u in lab_y]
+      if sum(inside) > 1
+        if (side==:right)
+          start == 0.0
+          stop = widths[i]
+        elseif (side==:left)
+          start = -widths[i]
+          stop = 0.0
+        elseif (side == :both)
+          start = -widths[i]
+          stop = widths[i]
+        end
+        lab_x[inside] = lab_x[inside] .+ linspace(start, stop, sum(inside)) .+ xcenter
+      end
+    end
+
+    append!(xp, lab_x)
+    append!(yp, lab_y)
+
+  end
+
+  x := xp
+  y := yp
+  seriestype := :scatter
+  ()
+
+end
+
+Plots.@deps beeswarm scatter

src/hist.jl

@@ -18,9 +18,8 @@ Plots.@deps density path
 # ---------------------------------------------------------------------------
 # cumulative density
 
-@recipe function f(::Type{Val{:cdensity}}, x, y, z; trim=false,
-                   npoints = 200)
-    newx, newy = violin_coords(y, trim=trim)
+@recipe function f(::Type{Val{:cdensity}}, x, y, z; trim::Bool=false, npoints=200)
+    newx, newy = violin_coords(y, trim=trim, n=npoints)
 
     if Plots.isvertical(d)
         newx, newy = newy, newx

src/violin.jl

@@ -1,11 +1,20 @@
-
 # ---------------------------------------------------------------------------
-# Violin Plot
+# Utility functions
 
 const _violin_warned = [false]
 
-function violin_coords(y; trim::Bool=false)
-    kd = KernelDensity.kde(y, npoints = 200)
+"""
+**Use kde to return an envelope for the violin and beeswarm plots**
+
+    violin_coords(y; trim::Bool=false, n::Int64=200)
+
+- `y`: points to estimate the distribution from
+- `trim`: whether to remove the extreme values
+- `n`: number of points to use in kde (defaults to 200)
+
+"""
+function violin_coords(y; trim::Bool=false, n::Int64=200)
+    kd = KernelDensity.kde(y, npoints = n)
     if trim
         xmin, xmax = Plots.ignorenan_extrema(y)
         inside = Bool[ xmin <= x <= xmax for x in kd.x]
@@ -14,8 +23,29 @@ function violin_coords(y; trim::Bool=false)
     kd.density, kd.x
 end
 
+"""
+**Check that the side is correct**
+
+    check_side(side::Symbol)
+
+`side` can be `:both`, `:left`, or `:right`. Any other value will default to
+`:both`.
+"""
+function check_side(side::Symbol)
+    if !(side in [:both, :left, :right])
+        warn("side (you gave :$side) must be one of :both, :left, or :right")
+        side = :both
+        info("side set to :$side")
+    end
+    return side
+end
+
+# ---------------------------------------------------------------------------
+# Violin plot
+@recipe function f(::Type{Val{:violin}}, x, y, z; trim=false, side=:both)
+
+    side = check_side(side)
 
-@recipe function f(::Type{Val{:violin}}, x, y, z; trim=true, side=:both)
     xsegs, ysegs = Segments(), Segments()
     glabels = sort(collect(unique(x)))
     bw = d[:bar_width]