Graphing Data

Table of contents:

Problem

You want to convert a bunch of data into a graph; usually a line chart, bar chart, or pie chart.

Solution

Use the Gruff library, written by Geoffrey Grosenbach. Install the gruff gem and build a Gruff object corresponding to the type of graph you want (for instance, Gruff::Line, Gruff::Bar,or Gruff::Pie). Add a dataset to the graph by passing data a label and an array of data points.

Heres code to create a graph that compares the running times of different sorts of algorithms:

	require 
ubygems
	require gruff

	g = Gruff::Line.new(600) # The graph will be 600 pixels wide.
	g.title = Algorithm running times
	g.theme_37signals # The best-looking theme, in my opinion.

	range = (1..101)
	g.data(Constant, range.collect { 1 })
	g.data(O(log n), range.collect { |x| Math::log(x) / Math::log(2) })
	g.data(O(n), range.collect { |x| x })
	g.data(O(n log n), range.collect { |x| x * Math::log(x) / Math::log(2) })

	g.labels = {10 => 
=10, 50 => 
=50, 100 => 
=100 }
	g.write(algorithms.png)

Figure 12-3 shows the graph it produces.

Figure 12-3. A line chart

Heres code to create a pie chart (shown in Figure 12-4). Note that the numbers given for the datasets don have to add up to 100. Gruff automatically scales the the pie chart to display the right proportions.

	p = 
Gruff::Pie.new
	p.theme_monochrome
	p.title = "Survey: the value of pi"
	p.data("About three", [3])
	p.data(3.14, [8])
	p.data(3.1415, [11])
	p.data(22/7, [8])

	p.write(pipie.png)

Figure 12-4. A pi chart

Discussion

Most of the time, programmers who need a graphing library need a simple graphing library: one that lets them easily produce a quick pie, line, or bar graph. Gruff works well for graphing simple datasets, but it doesn have the functionality of a fullfledged math program.

Gruffs interface for customizing the display of datasets also leaves something to be desired. Instead of letting you tweak the colors individually, it provides a number of themes that package together a background image, a text color, and a number of colors used in the graphs. Unfortunately, most of the provided themes are ugly (theme_37signals is pretty nice, though).

Heres a custom theme that makes monochrome graphs whose "colors" can be fairly easily distinguished. It takes advantage of the fact that its easy to distinguish dark shades of gray from light shades, and that lighter shades are more easily distinguishable from one another. The graphs in this recipe were actually created with this theme_monochrome, so that the "colors" would be more easily distinguishable in a printed book.

	class 
Gruff::Base
	 def theme_monochrome
	 reset_themes
	 @colors = "6E9C7ADB".scan(/./).collect { |c| "##{c * 6}"}
	 @marker_color = lack
	 @base_image = render_gradiated_background(white, white)
	 end
	end

This code adds writer methods for the various colors, letting you modify the current theme on an ad hoc basis. colors sets the colors used to differentiate datasets from each other. marker_color method sets the color of the title and axis labels. background sets the background to a solid color, or to a gradient between two colors.

	class 
Gruff::Base
	 def colors=(colors)
	 @colors = colors
	 end

	 def marker_color=(color)
	 @marker_color = color
	 end

	 def background=(color1, color2=nil)
	 color2 ||= color1
	 @base_image = render_gradiated_background(color1, color2)
	 end
	end

Problem

Solution

Figure 12-3. A line chart

Figure 12-4. A pi chart

Discussion

See Also