Example jgraph scripts

The unix utility jgraph is the best available tool for massaging data into two-dimensional graphs. The scripts describe the graphs at a fairly low level, however, and some experience is required to get the output you want. In this page, I've collected some jgraph scripts I've written for different purposes.

Some general things. This is not a basic introduction to jgraph. See

The jgraph manual.
James S. Plank's jgraph home page.

I find it useful to format jgraph scripts carefully, as can be seen in the examples on this page. I also find it useful to override most of the defaults, such as axis length, font, and so on.

In this ``log'' graph (source, data file 1, data file 2), a logarithmic scale is used on the y axis. Note some of the features on this graph:

Two lines (curves) can be distinguished using linetype but sometimes lines that are dashed (dotted, whatever) just look tacky. In this curve, I've used linethickness and gray. Remember that gray 0 is black and gray 1 is white.
The points to be plotted can be included directly in the script, but in this case I have used the shell command facility to all awk to suck them in from files. For this graph, either method would do (see this source for the same curves without awk), but in general use of shell commands makes scripting much easier. For example, several curves can be plotted from the one data file, by using awk to pull out different columns.
I haven't used the facility for placing a label under the graph. For tex/latex, labels are provided by the \caption command.

In this multi-line graph (source, data file), several lines are plotted from the one data file.

awk has been used to extract several sets of points from one data file. In one case, some of the data points are out of range; this is used by including a condition in the awk script.
A mix of strategies have been used to distinguish sets of lines from each other.
Font size is managed where the curves are specified rather than in the legend defaults. (And a more attractive font has been chosen.)
Note the use of precision 1 to get additional digits of precision on the hash marks on the y axis.

In this awk-based graph (source, data file, data file), awk is used in a slightly richer way; it filters the data, printing an average every tenth point instead of printing every point.

A graph with two axes (source) is useful for plotting two curves that have the same domain but different ranges.

A second newgraph is needed to draw the second y axis; a nodraw is used to make sure the xaxis isn't drawn twice.
draw_at is used to relocate the second y axis.
The legends have to be placed by hand. As they come from separate graphs, they are not neatly stacked together.

There are two ways to place text on a graph. One is with newstring, as in

newstring x 3.3 y 8.5 fontsize 9 : Upper

Another is with marktype, as in this graph with strings (source), and with horizontal and vertical delimiting regions.

Error bars

A graph with error bars (source, data 1, data 2, data 3, data 4, awk script) and points. Note the shifting of points to left and right to separate out the points shown at each integer level.

Bar graphs

Another useful form of graph is a bar graph (source). As this example shows, generating the bars is slightly tricky: for each bar, it is necessary to choose width and placement. However, the bars follow a simple, regular pattern, so this is easy to do with a script.

A variant of this graph is to place text labels (source) on the x axis.

Yet another variant of this graph is to use long labels (source) that have to be rotated to fit. These can be done, with some difficult, with newstring. The rotation is anti-clockwise, about the centre of the string. Thus, since the intention is to place the right-hand end of the string at a particular point, the location of the middle of each string has to be found individually. (It is not clear why this was done this way, when the vjr directive could have been used.)

A more principled way of doing this is with the generic hash_labels formatting directive, as in this version of the above bargraph (source).

Bar graphs can become very rich, as shown in this complex example. The source was built using a simple bash script using the two input files manybars.data and manybars.in.

Other stuff

It can be useful to add gridlines, but annoying (sometimes) to have them at the major hash marks. They can be easily added as newcurves; for example, use

newcurve pts 0 50 100 50 gray 0.7 linewidth 0.3 marktype none linetype dashed

to add a faint horizontal line at height 50 on a graph with domain [0,100]. While it is tiresome to add a newcurve instruction for every gridline, it is straightforward to generate them with an awk script.

Return to Justin Zobel's home page.