8 Stars
Updated Last
2 Years Ago
Started In
March 2013


Julia interface to Google Chart Tools.

A Google chart involves basically four steps:

  • a specification of a Google "DataTable"
  • a specification of chart options
  • a call to make the type of chart desired.
  • a call to draw the chart

This package allows this to be done within julia by

  • mapping a DataFrame object into a Google DataTable.

  • mapping a Dict of options into a JSON object of chart options. Many of these options can be specified through keyword arguments.

  • providing various constructors to make the type of chart

  • providing a method to see the charts. This is called through Julia's show mechanism. In general, the render method can draw the chart or charts to an IOStream or file.

A basic usage (see the test/ directory for more)

using GoogleCharts, DataFrames

scatter_data = DataFrame(
    Age    = [8,  4,   11, 4, 3,   6.5],
    Weight = [12, 5.5, 14, 5, 3.5, 7  ]

options = Dict(:title => "Age vs. Weight comparison",
           :hAxis =>  Dict(:title => "Age", 
                       :minValue => 0, 
                       :maxValue => 15),	
           :vAxis =>  Dict(:title => "Weight", 
                       :minValue => 0, 
                       :maxValue => 15)

scatter_chart(scatter_data, options)

For non-nested options, keyword arguments can be given, as opposed to a dictionary:

chart = scatter_chart(scatter_data, title="Age vs. Weight comparison")

There are constructors for the following charts (cf. Charts Gallery)

       area_chart, bar_chart, bubble_chart, candlestick_chart, column_chart, combo_chart,
       gauge_chart, geo_chart, line_chart, pie_chart, scatter_chart, stepped_area_chart,
       table_chart, tree_chart, annotated_time_line, intensity_map, motion_chart, org_chart,

The helper function help_on_chart("chart_name") will open Google's documentation for the specified chart in a local browser.

The names of the data frame are used by the various charts. The order of the columns is important to the charting tools. The "Data Format" section of each web page describes this. We don't have a mechanism in place supporting Google's "Column roles".

The options are specified through a Dict which is translated into JSON by JSON.to_json. There are numerous options described in the "Configuration Options" section of each chart's web page. Some useful ones are shown in the example to set labels for the variables and the viewport. Google charts seem to like integer ranges in the viewports by default. Top-level properties, can be set using keyword arguments.

In the tests/ subdirectory is a file with implementations with this package of the basic examples from Google's web pages. Some additional examples of configurations can be found there.

The GoogleCharts.render method can draw a chart to an IOStream, a specified filename, or (when used as above) to a web page that is displayed locally. One can specify more than one chart at a time using a vector of charts.

A Plot function

There is a Plot function for plotting functions with a similar interface as Plot's plot function:

Plot(sin, 0, 2pi)

A vector of functions:

Plot([sin, u -> cos(u) > 0 ? 0 : NaN], 0, 2pi, 
	   title="A function and where its derivative is positive",
           vAxis=Dict(:minValue => -1.2, :maxValue => 1.2))

The Plot function uses a line_chart. The above example shows that NaN values are handled gracefully, unlike Inf values, which we replace with NaN.

Plot also works for paired vectors:

x = linspace(0, 1., 20)
y = rand(20)
Plot(x, y)			         # dot-to-dot plot
Plot(x, y, curveType="function")         # smooths things out

parametric plots

Passing a tuple of functions will produce a parametric plot:

Plot((x -> sin(2x), cos), 0, 2pi)

scatter plots

The latter shows that Plot assumes your data is a discrete approximation to a function. For scatterplots, the Scatter convenience function is given. A simple use might be:

x = linspace(0, 1., 20)
y = rand(20)
Scatter(x, y)

If the data is in a data frame format we have a interface like:

using RDatasets
mtcars = dataset("datasets", "mtcars")
Scatter(:WT, :MPG, mtcars)

And we can even use with groupby objects:

iris = dataset("datasets", "iris")
d=iris[:, [2,3,6]]          ## in the order  "x, y, grouping factor"
gp = groupby(d, :Species)
Scatter(gp)                 ## in R this would be plot(Sepal.Width ~ Sepal.Length, iris, col=Species)
                            ## or ggplot(iris, aes(x=Sepal.Length, y=Sepal.Width, color=Species)) + geom_point()

Surface plots

Some experimental code is in place for surface plots. It needs work. The basic use is like:

surfaceplot((x,y) -> x^2 + y^2, linspace(0,1,20), linspace(0,2,20))

The above does not seem to work in many browsers and does not work reliably in IJulia (only success has been with Chrome).


The googleVis package for R does a similar thing, but has more customizability. This package should try and provide similar features. In particular, the following could be worked on:

  • Needs a julian like interface,
  • some features for interactive usage,
  • some integration with local web server.