```{js, echo=FALSE} $('.title').css('color', 'darkred')
```r library(knitr) Rpackage <- function (pkg) {sprintf("*%s*", pkg)} CRANpkg <- function (pkg){ cran <- "https://CRAN.R-project.org/package" fmt <- "[%s](%s=%s)" Rpackage(sprintf(fmt, pkg, cran, pkg)) } library(googleVis) options(prompt = "R> ", digits = 4, show.signif.stars = FALSE, gvis.plot.tag='chart')
In 2006 Hans Rosling gave an inspiring talk at TED [@HansRoslingTedTalk] about social and economic developments in the world over the past 50 years, which challenged the views and perceptions of many listeners. Rosling had used extensive data analysis to reach his conclusions. To visualise his talk, he and his team at Gapminder [@Gapminder] had developed animated bubble charts, aka motion charts.
Rosling's presentation popularised the idea and use of interactive charts. One year later the software behind Gapminder was bought by Google and integrated as motion charts into their Google Charts API [@GoogleVisApi], formerly known as Google Visualisation API.
In 2010 Sebastián Pérez Saaibi [@Saaibi2010] presented at the
R/Rmetrics Workshop on Computational Finance and Financial
Engineering, the idea to use Google motion charts to visualise R
output with the r CRANpkg("R.rsp")
package [@Rsp].
Inspired by those talks and the desire to use interactive data
visualisation tools to foster the dialogue between data analysts and
others the authors of this vignette started the development of the
r CRANpkg("googleVis")
package
[@RJournal:2011-2:GesmannCastillo] in August 2010.
The Google Charts API [@GoogleVisApi] allows users to create interactive charts as part of Google documents, spreadsheets and web pages. In this text, we will focus on the usage of the API as part of web pages.
The Google Public Data Explorer [@GooglePublicData] provides a good example, demonstrating the use of interactive charts and how they can help to analyse data.
The charting data can either be embedded into the HTML file or read
dynamically. The key to the Google Charts is that the data is
structured in a DataTable [@DataTable], and this is where the googleVis
package helps, as it transforms R data frames into JSON [@json]
objects, using the r CRANpkg("jsonlite")
package [@jsonlite], as
the basis for a DataTable.
As an example we shall look at the html-code of a motion chart from Google's visualisation gallery [@GoogleMotionChart].
1 <html> 2 <head> 3 <script type="text/javascript" 4 src="https://www.google.com/jsapi"> 5 </script> 6 <script type="text/javascript"> 7 google.load('visualization', '1', 8 {'packages':['motionchart']}); 9 google.setOnLoadCallback(drawChart); 10 function drawChart() { 11 var data=new google.visualization.DataTable(); 12 data.addColumn('string', 'Fruit'); 13 data.addColumn('date', 'Date'); 14 data.addColumn('number', 'Sales'); 15 data.addColumn('number', 'Expenses'); 16 data.addColumn('string', 'Location'); 17 data.addRows([ 18 ['Apples',new Date(1988,0,1),1000,300,'East'], 19 ['Oranges',new Date(1988,0,1),1150,200,'West'], 20 ['Bananas',new Date(1988,0,1),300,250,'West'], 21 ['Apples',new Date(1989,6,1),1200,400,'East'], 22 ['Oranges',new Date(1989,6,1),750,150,'West'], 23 ['Bananas',new Date(1989,6,1),788,617,'West'] 24 ]); 25 var chart=new google.visualization.MotionChart( 26 document.getElementById('chart_div')); 27 chart.draw(data, {width: 600, height:300}); 28 } 29 </script> 30 </head> 31 <body> 32 <div id="chart_div" 33 style="width:600px; height:300px;"> 34 </div> 35 </body> 36 </html>
The code and data are processed and rendered by the browser and is not submitted to any server^[Data Policy].
You will notice that the above HTML code has five generic parts:
DataTable
(ll. 11 - 24),<div>
element to add the chart to the page (ll.
32 -- 34).These principles hold true for most of the interactive charts of the Google Chart Tools, see the examples in the next section.
However, before you use the API you should read the Google Terms of Service [@GoogleTerms].
The googleVis package provides an interface between R and the Google Chart Tools. The functions of the package allow the user to visualise data stored in R data frames with Google Charts.
The output of a googleVis function is HTML code that contains the
data and references to JavaScript functions hosted by Google. A
browser with an Internet connection is required to view
the output, and for a very few chart types, notably motion charts, also Flash.
Examples of several chart types are shown below, which have been combined with
the gvisMerge
function.
## Code for screen shot library(googleVis) df <- data.frame(country=c("A", "B", "C"), val1=c(1,3,4), val2=c(23,12,32)) Line <- gvisLineChart(df, xvar="country", yvar=c("val1", "val2"), options = list( width=300, height=300, legend = "top", title="Hello World", titleTextStyle = "{color:'red',fontName:'Courier', fontSize:16}", curveType='function')) ATL <- gvisAnnotationChart(Stock, datevar="Date", numvar="Value", idvar="Device", titlevar="Title", annotationvar="Annotation", options=list(displayAnnotations=TRUE, legendPosition='newRow', width=300, height=300) ) Gauge <- gvisGauge(CityPopularity, options=list(min=0, max=800, greenFrom=500, greenTo=800, yellowFrom=300, yellowTo=500, redFrom=0, redTo=300, width=300, height=200)) Geo <- gvisGeoChart(Exports, locationvar='Country', colorvar='Profit', options=list(projection="kavrayskiy-vii", width=300,height=200)) # Datetime example dat <- data.frame(Room=c("Room 1","Room 2","Room 3"), Language=c("English", "German", "French"), start=as.POSIXct(c("2014-03-14 14:00", "2014-03-14 15:00", "2014-03-14 14:30")), end=as.POSIXct(c("2014-03-14 15:00", "2014-03-14 16:00", "2014-03-14 15:30"))) TL <- gvisTimeline(data=dat, rowlabel="Language", start="start", end="end", options=list(width=300, height=200)) Tree <- gvisTreeMap(Regions, "Region", "Parent", "Val", "Fac", options=list(width=300, height=200, fontSize=16, minColor='#EDF8FB', midColor='#66C2A4', maxColor='#006D2C', headerHeight=20, fontColor='black', showScale=TRUE)) M <- gvisMerge( gvisMerge(gvisMerge(Line, ATL, horizontal = TRUE, tableOptions="cellspacing=10"), gvisMerge(Gauge, Geo, horizontal =TRUE, tableOptions="cellspacing=10")), gvisMerge(TL, Tree, horizontal = TRUE, tableOptions="cellspacing=10")) plot(M)
You can install googleVis in the usual way from CRAN, e.g.:
install.packages('googleVis')
The installation was successful if the
command library(googleVis)
gives you the following message:
library(googleVis)
library(googleVis)
cat(googleVis:::gvisWelcomeMessage())
The individual functions of the googleVis package are documented in the help pages. Here we will cover only the basic concepts of the package.
As an example, we will show how to generate a geo chart. It works
similarly for the other APIs. Further examples are covered in the demos^[See demo(package="googleVis")
for a list of the available demos] of the googleVis
package.
The design of the visualisation functions is fairly generic. The name
of the visualisation function is 'gvis' + ChartType
. So for
a geo chart we have:
gchart <- gvisGeoChart(data, locationvar = "", colorvar = "", sizevar = "", hovervar = "", options = list(), chartid)
Here data
is the input data.frame
, locationvar
,
colorvar
, sizevar
and hovervar
specify the various columns
used for the plot. Display options are set in an
optional list, which we discuss in more detail later.
The options and data requirements
follow those of the Google Charts API and are documented in the
help pages, see
help('gvisGeoChart')
The argument chartid
allows the user to set a chart ID of the
output chart manually. If the argument is missing a random ID using
tempfile(pattern='')
will be generated. Unique chart IDs are required to place more than
one chart on a web page.
The output of a googleVis function is a list of lists (a nested list) containing information about the chart type, chart ID and the HTML code in a sub-list with header, chart, caption and footer.
The idea behind this concept is that users can get a complete web page, while at the same time offer a facility to extract specific parts, such as the chart itself. This is particularly helpful if the package functions are used in solutions where the user wants to feed the visualisation output into other sites.
The output of a googleVis function will be of class gvis
and list
. Generic print (print.gvis
) and plot
(plot.gvis
) functions exist to ease the handling of such objects.
To illustrate the concept we shall create a Geo chart using the
Exports
data set.
Following the documentation of the Google Geo Chart API we need a data set which has at least one column with the location variable.
As an example we use the Exports
data set:
data(Exports)
Exports
Here we will use the columns 'Country'
and
'Profit'
as location and colour variable respectively.
gchart <- gvisGeoChart(data = Exports, locationvar='Country', colorvar='Profit', options=list(projection="kavrayskiy-vii", width=400, height=200))
The structural output of gvisGeoChart
is a list of lists as
described below.
str(gchart)
## This statement avoids truncation cat(paste(substring( capture.output( str(gchart) ) , 0, 66), sep="\n", collapse="\n"))
The first two items of the list contain information about the chart type used and the individual chart ID:
gchart$type gchart$chartid
The html output is a list with header, chart, caption and footer. This allows the user to extract only certain parts of the page, or to create a complete html page.
The header part of the html page has only basic html and formatting tags:
print(gchart, tag='header')
Here we used the print
statement with the tag
'header'
instead of gchart$html$header
to achieve
a formatted screen output. This is the same output as
cat(gchart$html$chart)
.
The actual Google visualisation code is stored with the data
as a named character vector in the chart
item of the HTML
list. The chart is made up of several JavaScript and HTML
statements. Please notice that the JavaScript functions are
uniquely named with the information of the chart ID.
This concept allows the user get all the chart code directly or only
specific parts; see the examples in the help page of print.gvis
for more details.
names(gchart$html$chart)
The complete chart can be displayed via:
print(gchart, tag='chart') ## or cat(gchart$html$chart)
Similarly you can also access specific components of the chart, e.g. (output truncated)
cat(gchart$html$chart['jsChart']) # or print(gchart, 'jsChart')
cat(paste(substring( capture.output( cat(gchart$html$chart['jsChart']) ) , 0, 66), sep="\n", collapse="\n"))
A basic chart caption and html footer are the final items of the html list (output truncated):
print(gchart, tag='caption')
cat(paste(substring( capture.output( cat(gchart$html$caption) ) , 0, 66), sep="\n", collapse="\n"))
print(gchart, tag='footer')
cat(paste(substring( capture.output( cat(gchart$html$footer) ) , 0, 66), sep="\n", collapse="\n"))
To display the page locally, simply type:
plot(gchart) # returns invisibly the file name
The plot method for gvis
-objects creates HTML files in a temporary
folder using the type and chart ID information of the object and it will display the output using the R built-in web server.
Note that the chart caption provides a link to the chart code via the chart ID for copy and paste.
The R command tempdir()
will show you the path of the
per-session temporary directory, in which the files were written.
You can write the chart into a local html file via the print
command with the file argument, e.g.
print(gchart, file="myGoogleVisChart.html")
Alternatively use the function plot.gvis
explicitly,
e.g. suppose your html file is stored in
/Users/JoeBloggs/myGoogleVisChart.html
. Using the
plot.gvis
the file will be copied into a temporary directory
and displayed via the R HTTP help server with, in the same way as a
gvis-object:
plot.gvis("/Users/JoeBloggs/myGoogleVisChart.html")
Setting the various options of a googleVis objects can be a bit
cumbersome at first. The options follow those of the Google
Charts API and can be set via a named list using the argument options
.
In the following example, we create a line chart and set various
options:
df <- data.frame(country=c("US", "GB", "BR"), val1=c(1,3,4), val2=c(23,12,32)) Line <- gvisLineChart(df, xvar="country", yvar=c("val1","val2"), options=list( title="Hello World", titleTextStyle="{color:'red', fontName:'Courier', fontSize:16}", backgroundColor="#D3D3D3", vAxis="{gridlines:{color:'red', count:3}}", hAxis="{title:'Country', titleTextStyle:{color:'blue'}}", series="[{color:'green', targetAxisIndex: 0}, {color: 'orange',targetAxisIndex:1}]", vAxes="[{title:'val1'}, {title:'val2'}]", legend="bottom", curveType="function", width=500, height=300 )) plot(Line)
As you can see from the example above, the simpler options can be set by name=value
, e.g. width=500
, while the more complex options with sub-components are
listed in curly brackets {}
and arrays []
, e.g. to define the two
axes.
Generally, the following rules apply:
options=list(width=200, height=300)
. Boolean arguments are set to
either TRUE
or FALSE
, using the R syntax.color
, and are wrapped in [ ]
, e.g.
options=list(colors="['#cbb69d', '#603913', '#c69c6e']")
parameter:value
. Boolean values
have to be stated as 'true'
or 'false'
.
For example the Google documentation states the formatting options for the
vertical axis as vAxis.format
.
Then this parameter can be set in R as:
options=list(vAxis="{format:'#,###%'}")
. titleTextStyle.color
, titleTextStyle.fontName
and
titleTextStyle.fontSize
, then those can be combined in one list item
such as: options=list(titleTextStyle="{color:'red',fontName:'Courier', fontSize:16}")
[ ]
. For example to set the labels for left and right axes use:
options=list(vAxes="[{title:'val1'}, {title:'val2'}]")
A special option for all charts is gvis.editor
, which adds an
edit button to the page, allowing the user to edit, change and
customise the chart on the fly. The content of the option gvis.editor
describes the label of the browser button.
Editor <- gvisLineChart(df, options=list(gvis.editor='Edit me!')) plot(Editor)
Suppose you have an existing web page and would like to integrate the
output of a googleVis function, such as gvisLineChart
.
In this case you only need the chart output from
gvisLineChart
. So you can either copy and paste the output
from the R console
print(gchart, 'chart') ## or cat(gchart$html$chart)
into your existing html page, or write the content directly into a file
print(gchart, 'chart', file='myfilename')
and process it from there.
Shiny is a package by RStudio, which makes it incredibly easy to build interactive web applications with R.
With version 0.4.0 of googleVis support for shiny
apps was added. Joe Cheng contributed the renderGvis
function which allows users to use googleVis output in shiny in a
similar way to other plotting functions.
The following example has been taken from the help file of
renderGvis
. It displays a scatter chart where the user can
select the data set to be displayed.
# server.R library(googleVis) shinyServer(function(input, output) { datasetInput <- reactive({ switch(input$dataset, "rock" = rock, "pressure" = pressure, "cars" = cars) }) output$view <- renderGvis({ gvisScatterChart(datasetInput()) }) }) # ui.R shinyUI(pageWithSidebar( headerPanel("googleVis on Shiny"), sidebarPanel( selectInput("dataset", "Choose a dataset:", choices = c("rock", "pressure", "cars")) ), mainPanel( htmlOutput("view") ) ))
You can run the example locally with the following statement.
library(shiny) runApp(system.file("shiny/", package="googleVis"))
print.gvis
and plot.gvis
In googleVis version 0.3.2 the function plot.gvis
gained the
same argument as print.gvis
: tag
. By default the
tag
argument is set to NULL
in plot.gvis
and the plot function will display its output in a browser
window. However, if tag
is not NULL
the function
plot.gvis
will behave exactly like print.gvis
.
The default tag
can be set for both functions globally via
the options()
function. On package load googleVis sets
options(gvis.print.tag='html')
and
options(gvis.plot.tag=NULL)
.
Suppose you would set options(gvis.plot.tag='chart')
then all
following plot statements would print the chart part of the
gvis-object only, without opening a browser window. This might
seem a bit odd at first, yet it becomes helpful when you write R
Markdown files for knitr
or files for other packages such as
R.rsp
.
While you draft your file you may want to see the output of googleVis
in an interactive way, so you set options(gvis.plot.tag=NULL)
at the top of the file and you change the setting to 'chart'
before you parse the file, say with knitr
. This will
ensure that all plot statements return the HTML code of the chart,
rather than opening browser windows.
Using googleVis with knitr [@knitr] is a convenient way of creating interactive reproducible reports. The approach taken by knitr is similar to Sweave, you can combine R code with text and formatting tags. However, knitr can also export to HTML, which is required to embed googleVis charts.
To include googleVis output into a knitr document you have to set
the self_contained
option to false
in the YAML header:
--- title: "My document" output: html_document: self_contained: false ---
Furthermore, the chunk option results
should be set to 'asis'
,
so that the html output is written into the markdown file.
You can either use the print
statement:
cat(paste(sep = "\n", "```r", "gchart <- gvisColumnChart(CityPopularity, 'City', 'Popularity',", " options=list(width=550, height=450,", " legend='none'))", "print(gchart, 'chart')", "```" ))
or alternative change the behaviour of the plot
function via setting
options(gvis.plot.tag = 'chart')
. With this setting plot(x)
will no longer open a browser window, but produce the same output as
print(x, tag='chart')
, if x
is a gvis-object.
Hence, setting the option gvis.plot.tag
in a knitr markdown
Rmd-file to 'chart'
will automatically turn all following
plot statements into html output.
Note that you can use the options()
command in your knitr
file to switch between an interactive mode, where you are likely to
experiment, via copying and pasting R code into the console and
running knit
on the whole file.
A more comprehensive example is given in the help file to ?plot.gvis
.
The Google Chart Tools are designed for web pages, so it should be no surprise that it can be difficult or impossible to embed googleVis output in traditional presentation software like MS PowerPoint, Google Docs, OpenOffice Impress or Apple Keynote.
The easiest way is to include screen shots into the slide with links to the live web pages. But this approach requires the presenter to switch between applications during her talk. This can be fun, but quite often it is not.
An alternative would be to build the presentation as a web page itself.
A popular approach here is the slidify
package by Ramnath Vaidyanathan,
[@slidify] that builds on the knitr Markdown approach of the previous
section. An example of a slidify
presentation is the googleVis
tutorial given at the useR! conference in 2013, @googleVisTutorial.
A trendline is a line superimposed on a chart revealing the overall direction of the data. Google Charts can automatically generate trendlines for Scatter Charts, Bar Charts, Column Charts and Line Charts.
For more details visit: https://developers.google.com/chart/interactive/docs/gallery/trendlines
read_demo('Trendlines', 'googleVis')
Roles add the ability to pass columns for further processing downstream. Role columns must follow column they pertain to. Proper naming conventions must be observed. For example, roles fulfilling tooltip roles and must be called "foo.blah.tooltip". For more details see the Google documentation.
The following examples should help to illustrate the concept.
read_demo('Roles', 'googleVis')
The first example uses a data set that has the additional column
pop.html.tooltip
with the first 11 letters of the Latin alphabet.
This column is mapped automatically as a tooltip when the user hovers
over the chart point.
HTML code can be embedded into the tooltip as well, if the option
isHtml
is set to true.
Often it is helpful to highlight certain parts of the data. The Google API distinguishes between certainty and emphasis. In a similar way to above additional columns with boolean values have to be added to the data.
Using roles with column or bar charts has some specifics. Instead of 'emphasize' use 'style' to change the colours.
Setting the annotations style to 'line' allows adding little reference lines to the plot.
Intervals help to add error bars, confidence levels, etc.
Note that the options are set either via interval
or intervals
,
if set to all intervals. The examples below give an indication of what
can be achieved with intervals.
For more details visit the Google documentation.
The tool tips in geo charts need a little work around, a Tooltip.header
variable has to be set to an empty string:
In this section we present ideas which go beyond the usual coding in R and are somewhat experimental.
Google visualisations can fire and receive events. It exposes the following two JavaScript methods:
google.visualization.events.trigger()
fires an event,google.visualization.events.addListener()
listens for events.Here is an example of registering to receive the selection event from the Google documentation:
var table = new google.visualization.Table(document.getElementById('table_div')); table.draw(data, options); google.visualization.events.addListener(table, 'select', selectHandler); function selectHandler() { alert('A table row was selected'); }
We will only deal with this special case of a 'select' event of the 'addListner' method. This event is available for most visualisations and acts on user interactions, e.g. user selection clicks.
The 'addListener' method expects JavaScript code, which can be embedded
into a gvis-object via options
as (undocumented) parameter
gvis.listener.jscode
.
Here are some examples:
Look up the selected item in Wikipedia:
jscode <- "window.open('https://en.wikipedia.org/wiki/' + data.getValue(chart.getSelection()[0].row,0)); " J1 <- gvisGeoChart(Exports, locationvar='Country', sizevar = 'Profit', options=list(dataMode="regions", gvis.listener.jscode=jscode)) plot(J1)
In the same way we can use the code in other charts, e.g. org- or line charts:
plot(gvisOrgChart(Regions, options=list(gvis.listener.jscode=jscode))) plot(gvisLineChart(Regions[,c(1,3)], options=list(gvis.listener.jscode=jscode)))
In the following more advanced example the selected value of a table is displayed in a message box:
jscode <- " var sel = chart.getSelection(); var row = sel[0].row; var text = data.getValue(row,1); alert(text); " J2 <- gvisTable(Population[1:5,], options=list(gvis.listener.jscode=jscode)) plot(J2)
For more details see the demo(EventListener)
and
Google Charts Reference.
Other R packages provide alternatives to the Flash based Google motion chart, such as plotly, gganimate
No, not directly. The Google Charts API is designed for dynamic web output on your screen and not on paper.
For further details see Google's online documentation on printing PNG charts.
No, unfortunately not. The colours are set by the Google Charts API and cannot be changed by the user.
Note that Flash charts may not work when loaded as a local file due to security settings, and therefore require to be displayed via a web server. However, you can overcome this issue by changing your Flash security settings. Tony Breyal posted the following solution on stackoverflow.com:
Motion charts (also geo maps and annotated time lines) are rendered in your browser using Flash, unlike most other charts which use HTML5. Unfortunately, Flash is not directly supported on iOS devices such as iPads and iPhones.
Unfortunately, there are no arguments such as ylim
and xlim
.
Instead, the Google Charts axes options are set via hAxes
and
vAxes
, with 'h' and 'v' indicating the horizontal and vertical axis.
More precisely, we have to set viewWindowMode:'explicit'
and set the viewWindow
to the desired min
and max
values. Additionally, we
have to wrap all of this in [{}]
brackets as those settings are sub
options of h/vAxes
. There are also options minValue
and
maxValue
, but they only allow you to extend the axes ranges.
Here is a minimal example, setting the y-axis limits from 0 to 10:
dat <- data.frame(x=LETTERS[1:10], y=c(0, 4, -2, 2, 4, 3, 8, 15, 10, 4)) area1 <- gvisAreaChart(xvar="x", yvar="y", data=dat, options=list(vAxes="[{viewWindowMode:'explicit', viewWindow:{min:0, max:10}}]", width=500, height=400, title="y-limits set from 0 to 10"), chartid="area1ylim") plot(area1)
The googleVis package converts data frames into JSON objects. The column names of the resulting JSON tables are encapsulated with single speech marks.
Hence apostrophes in column names of your input data frame have to be encapsulated by a double backslash.
Here is a little example:
df <- data.frame("Year"=c('2009', '2010'), "Alice\\'s salary"=c(86.1, 93.3), "Bob\\'s salary"=c(95.3, 100.5), check.names=FALSE) gchart <- gvisColumnChart(df, options=list(vAxis='{baseline:0}', title="Salary", legend="{position:'bottom'}")) plot(gchart)
The charts have a lot of options which allow you to change the look and feel of the output, see the help files for more details. However, googleVis provides only an interface to the Google Charts API. If you have specific questions to the charts then please ask the Google Visualisation API newsgroup.
For frequent ask questions regarding the API check: https://developers.google.com/chart/interactive/faq.
Review the Google Terms of Service and get in touch with your colleagues in IT / Legal. If in doubt contact Google directly.
Should you find any issues or bugs with googleVis, then please drop us a line or add them to our issues list: (https://github.com/mages/googleVis/issues)
Please cite googleVis if you use it in your work or publications:
citation("googleVis")
\bibliographystyle{alpha} \bibliography{googleVis}
Any scripts or data that you put into this service are public.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.