the pocket guide to

gnuplot

V.H. Belvadi
A note from the author
This guide has curious but sincere origins. It was written following the publica-
tion of an article1 on installing gnuplot on macOS (then OS X), outlining a process
which was not as straightforward as it should have been. The article became unusu-
ally popular, at least in terms of how many other sites, especially academic ones,
linked back and sent traffic to it, and it soon became apparent that most people
interested in learning gnuplot could probably do with a handy guide dedicated to
helping them get started with the program.
Two other motivations exist for this: the first is my own learning of gnuplot for
no reason other than the fact that it excited me, and the second is that it was
part of the postgraduate course at my institution and I thought having an in-house
manual would probably make things convenient for students.
Time is a teacher. By the time this short guide had been completed the mean-
inglessness of an in-house manual in a connected world had become clear and the
book was simply made available publicly for anyone interested in making use of it.
It was designed to get you started with gnuplot and not to master it. In fact there
is no such thing as absolute mastery: if you can do what you set out to do with
gnuplot then you have mastered it in that scenario. But, sooner or later, you will
come across a highly specific requirement that makes you stop and question your
entire understanding of a subject.
This guide is extremely short for a reason: it was meant to be in your pocket as
a constant help while you started learning gnuplot and, like a couple of other intro-
ductory courses I have designed for the postgraduate level (e.g. C for physicists,
LATEX), the purpose of this book is to teach you enough gnuplot to give you a good
grasp of it, and enough knowledge, by any count, to start teaching yourself the rest
(and other similar programs). After all, where is the fun in not being self-taught
in at least some disciplines?

the pocke guide to gnuplot

First edition, 2015.

Second edition, 2017.

Typeset in Baskervald and Nimbus Sans.

This work is licensed under the Creative Commons

Attribution-ShareAlike 4.0 International License.
To view a copy of this license, please visit
creativecommons.org/licenses/by-sa/4.0.

E-mail the author at vh@belvadi.com with your

comments and suggestions or for clarifications.

1 https://vhbelvadi.com/installing-gnuplot-on-mac
What is gnuplot ?
Plotting and analysing experimental data is a major part of physics and gnuplot is
one of the most important, and among the oldest, tools used for this. It is a
command-line driven graphing application available on nearly all platforms and
can be used to draw extremely high quality precise graphs from both functions
and data points with extremely simple commands.
Download gnuplot and set it up on your system using the appropriate executable
from www.gnuplot.info. Installing on macOS is simpler if done via the Terminal.
For macOS follow this article: vhbelvadi.com/installing-gnuplot-on-mac/.
Needless to say you have to work on gnuplot on the Terminal. It can, however,
communicate with appropriate data and text files carrying space-separated tables.
It outputs as well (and not in the Terminal) to several formats: png, eps, pdf, jpeg,
svg etc. Therefore, even if an example in this guide uses a particular format you
can replace it with any other without trouble.
Note that throughout this guide code is printed in a monospaced font and
arguments (where custom data must be input based on your requirements) are
mentioned within hangle bracketsi so make sure that, while working in the Ter-
minal, you do not type the brackets themselves, just the text they contain.
Once it has been installed on your system, begin accessing gnuplot by simply
typing gnuplot on screen. It should take you to the gnuplot environment and start
up its default terminal (i.e. an interactive display environment) for you. To set
your own terminal use the set terminal hnamei command. where hnamei is the
terminal you wish to use. You can use x11 or aquaTerm. Try this command as
soon as you enter gnuplot :

set terminal postscript eps enhanced

Discussing the specific capabilities of each terminal beyond this point is outside
the scope of this guide. Once you are done working with gnuplot simply type exit
to quit gnuplot and return to your standard Terminal prompt, usually the dollar
prompt. By default gnuplot does not save outputs; this is partly to ensure test
renders are not saved over and over again. So make sure you save your output
with an appropriate name for later use (this will be explained below). Once the
output is ready you can use it like any other image inside any document, .pages,
.docx, .tex and others.

Plotting functions
Now that you have gnuplot set up and are in the gnuplot environment it is time to
draw your first plot. Type the following command and hit return/enter:

plot sin(x)

You should see the sine curve output on a separate window as shown in fig. 1
below. You can plot other functions as well, like cos(x), atan(x) etc. using the
plot hfunctioni command. The function you plot is of the form y = f (x) where

1
the plot command takes f (x) as its argument, which is why y = sin(x) is plotted
as plot sin(x) and y = x 2 is plotted as plot x^2 and so on.

Figure 1: Plotting sin(x).

Let us now plot the trajectory of a projectile and use that graph as our main object
to learn other gnuplot commands. Use the following command:

plot tan(pi/4)*x-((9.8/2*10*10*(cos(pi/4))**2)*x**2)

Observe that this is our standard projectile trajectory given by the equation

gx 2
y = tan(θ)x −
2u2 cos2 (θ)
where we have used a 45◦ angle of launch θ, in radian, and an initial velocity of
10 units. The use of two asterisks, as in x**2 implies indices, in this case x squared.
We have assumed the magnitude of the acceleration due to gravity is g = 9.8 ms−2
in our example. The plot function takes as its argument only the right-hand f (x)
side of the y = f (x) equation above. As expected we end up with a parabolic
trajectory like the output shown in fig. 2.

Refining outputs
Of course there are a few problems with our output in terms of its presentation.
The graph is perfect, but the legend/key, the axes etc. can use some improvements.
This is what the remainder of this section will focus on.
First let us start by getting rid of that equation at the top blocking our graph.
This is particularly useful if you have multiple lines as it serves as a legend or key.
In gnuplot commands must be given one after the other, often in any order, but

2
 Figure 2: Projectile trajectory.

before the plot (or equivalent) command is given. gnuplot keeps these settings in
memory for the current and subsequent plots.
To remove the legend use the command

unset key

or you can move the legend around to a more convenient location by specifying
that location, such as the top right or bottom centre:

set key top right

set key bottom center

Most such commands are given using the set keyword in gnuplot so they are easy
to recognise. Likewise unset undoes a setting. To reset all settings simply type

reset

Let us go a few steps further and label our axes, title our graph, draw a grid for
better coördinates and so on. In gnuplot type the following commands and hitting
return/enter after each. Nothing will happen visibly when you do this as gnuplot is
simply committing these new settings to memory. When you finally plot your data

3
or function you can see the result of these settings.

set xlabel "Distance (m)"

set ylabel "Height (m)"

set title "Projectile"

set zeroaxis

set grid

These five commands should be self-explanatory. They set the x-axis label, the
y-axis label, the graph title, display the (0,0) axes and display a graph/grid in the
background respectively. Once you do this run the plot the command again for
our trajectory equation and you should see an output that looks like fig. 3 below.

Figure 3: A better-looking projectile trajectory.

This is much better looking. The key is not obstructing our view, there is a
nice title and the axes are now labelled. But the exact numbers on the axis are
disorienting (to me anyway) so let us try to improve the precise numbering of axes.

4
First we set new ranges for our axes with the [minimum:maximum] syntax:

set yrange [0:30000]

set xrange [-10:10]

Notice that our graph starts at −25, 000 m so, as our final step, we need to offset
the entire graph up by that value to bring it to 0 m which we do by simply adding
25000 to f (x) as follows:

plot (tan(pi/4)*x-((9.8/2*10*10*(cos(pi/4))**2)*x**2))+25000

So far we have made some good improvements but the graph still does not let
us quickly check x and y values. Having more gridlines would help us; in other
words, we need to reduce the least count of our axes to, perhaps, 2 m on the x-
and 2,500 m on the y-axis. You should now see a more comprehensible output as
in fig. 4 below.

Figure 4: Improved ranges.

You can even switch scales from decimal (used here) to logarithmic with

set logscale x

set logscale y

5
for the x- and y-axis respectively. You can even add minor gridlines in-between
existing ones, say for every 0.5 m on the x- and 500 m on the y-axis in the graph
above, using the commands

set mxtics 0.5

set mytics 500

To quickly plot your last plot, instead of retyping the plot command, say

replot

You can even plot multiple plots on the same graph. To have some fun let us see
the effect of the launch angle on our projectile. To see the effect of launch angles
π/3 and π/6 on our projectiles we replace the angles appropriately in our equation
and then use the command to plot multiple functions by simply separating each
function with a comma as follows:

plot (tan(pi/4)*x-((9.8/2*10*10*(cos(pi/4))**2)*x**2))+25000,
(tan(pi/3)*x-((9.8/2*10*10*(cos(pi/3))**2)*x**2))+25000,
(tan(pi/6)*x-((9.8/2*10*10*(cos(pi/6))**2)*x**2))+25000

Let us also refine our axes and legend with the following commands:

set xrange [0:15]

set yrange [0:27000]

set key top right

and then replot to get the output shown in fig. 5.

Plotting from files

A lot of experimental set-ups output data as large sets of numbers. Such ordered
pairs can also be plot using gnuplot by simply mentioning the file name. Say we
have a file called data.dat with the contents shown below. The data is simply
separated by tabbed spaces and each visible column is interpreted as just that by
gnuplot so the spacing is all we need to pay attention to. You can use any text
editor, from TextEdit to VI to Notepad, to create such a data file.
Say we want to plot the first and second column against each other, we can do
so using the following command:

plot "data.dat" u 1:2 w linespoints

6
 Figure 5: Multiple plots in one graph.

Remember to first reset your gnuplot or x- and y-axis ranges from the previous plot
will be applied to this as well. You can alternately exit and return to gnuplot but
that is a little time consuming compared to the good old reset command.

0 0 4
2 3 5
4 4 7
6 5.75 11
8 9 16.25
10 10.5 22

data.dat

If you did it right you should have a plot that looks like fig. 6. There are two
new keywords in this plot command: the u stands for ‘using’ and the w stands for
‘with’ and the 1:2 tells gnuplot to plot the first column as the x-axis and the second
as the y-axis. You can also use the full keywords using and with instead of their
respective single-letter shorthands.
Also the keyword linespoints tells gnuplot to mark coördinates with points
and connect them with lines. You can just as well choose to have only points or
lines on your graph.

7
 Figure 6: Plotting from a file.

Fitting data
One observation to make is that fig. 6 is not a straight-line graph. What if we
wanted to take this data and fit it into a straight line? That too can be done in
gnuplot with a couple of simple commands. (The letter t is short for title; note how
the title is used in the key when plotting multiple plots on one graph.)

f(x) = m * x + c

fit f(x) "data.dat" u 1:2 via m,c

plot "data.dat" w points, m * x + c w lines t "Line-fit curve"

After you execute the fit command gnuplot shows you its fitting calculations and
the values it has arrived at for the slope (m) and intercept (c). This is what allows
us to use m * x c+ in the next command to plot a straight-line into which our
data fits. We can also fit our data into cubic splines and Bezier curves with:

plot "data.dat" u 1:3 smooth csplines title "Cubic spline",

"data.dat" u 1:3 w lines title "Raw",
"data.dat" u 1:3 smooth sbezier title "Bezier"

to arrive at the multi-plot in fig. 8. Note the subtle differences between the curves.

8
 Figure 7: Fitting data to a line.

Figure 8: From left to right: Bezier, raw and cubic spline fits.

When you are typing a command in gnuplot if you find the line becoming longer
than you like, use the backward slash ‘\’ character: type your command, hit \, hit
return/enter and continue typing. gnuplot interprets such lines as one command.

9
Saving files
The clever reader may have noticed by now that none of their carefully drawn plots
are being saved anywhere. This is by design. You can use gnuplot to keep refining
your plot as much as you want and when you are finally ready to save your finalised
work you can give it a name and save it:

set output "mygraph.eps"

Now plot your file or function and it will be saved into your current working direc-
tory as an .eps file2 . This only works if, as we said at the start of this guide, you
used the postscript eps enhanced terminal.
Alternately you can use the set terminal png or the set terminal gif com-
mands to enable .png and .gif outputs respectively. Also try the latex terminal to
get a picture environment code as your output which you can then simply copy
and paste into a LATEX file3 .
Be sure to enter your working directory before you start gnuplot because all your
outputs will start getting saved into that directory. Exiting gnuplot in-between will
reset all your old plot settings.
This guide should have whet your appetite about gnuplot . As stated in the
beginning, this is not an exhaustive guide but you should, by now, be in a position
to explore gnuplot all by yourself and discover several other capabilities. Start by
looking at, for example, colouring your graphs, and drawing three-dimensional
plots. And remember to always have fun.

* * *

Ever since the first edition of this self-study guide was published several readers wrote to me enquiring
about the code that would output a torus like the one on the cover page. Here it is for you to try:

set nokey
set parametric
set hidden3d
set view 30
set isosamples 30,20
splot [-pi:pi][-pi:pi] cos(u)*(cos(v)+3), sin(u)*(cos(v)+3), sin(v)

2 Encapsulated PostScript.
3 Such a picture environment code, while incredibly convenient, can become extremely lengthy.
For instance, plotting fig. 8 as a LATEX picture leaves you with 475 lines of code for a single plot.

10
Notes