Linux Goodies

In Pursuit Of The Perfect O/S



A Review of the Free Octave Math Language

Zazzle Linux Gifts

Octave, a Free Math Language Very Similar To MATLAB

Do you need a matrix language quick, and one that won't cost you an arm and a leg? One that's highly compatible with Matlab?

How about a language that offers all of that, is loaded with supplemental math routines, and can cost you -- nothing?

If that sounds enticing, then read on to learn about the popular matrix scripting language Octave.

If you like what you see and want more complete information, check out the GNU Octave Beginner's Guide.

The image below was made with the Octave mesh command. It shows what's called a sin(x)/x curve. It also happens to be the intensity graph of a star image as seen through a telescope. It always makes an impressive example for surface plot utilities.

Octave Mesh Plot

Octave Mesh Plot

Unlike the GUI interface of MATLAB, Octave has a simple command line interface (no GUI). But it does have a nice history mechanism allowing one to scroll back to, and if desired modify, previous commands. In fact, the history is maintained even when exiting Octave and re-entering it later.

In Windows, Octave isn't spectacular. It runs in kind of a pseudo-Linux environment, and this makes file access withing Windows an awkward and bit arcane procedure.

The version of Octave reviewed in this article was included in my Debian Etch Linux distribution, and thus was easily installed.

Octave Syntax

Octave is sort of the GNU answer to the commercial language MATLAB. That is, it is a scripting matrix language, and has a syntax that is about 95% compatible with MATLAB. It's a language designed by engineers, and thus is heavily loaded with routines commonly used by engineers. It has many of the same time series analysis routines, statistics routines, file commands, and plotting commands of the MATLAB language.

Octave even uses the "m" file method of creating functions. That is, user created functions are stored in files with an m extension, and can be called in Octave programs or interactively without explicitly loading them. To a very large degree, Octave can run MATLAB m files, and vice versa.

Octave's math routines only work on scalars, and 1 or 2 dimensional matrices. One can make data structures of more than 2 dimensions, such as a structure that is a collection of 2 dimensional matrices in a 3 dimensional stack, but math routines can only be applied to the two dimensional elements of the data structure.

As is true with MATLAB, Octave math operators default to matrix operations. For example, the following expression does a full matrix multiply of matrices x and y:

z = x * y;

If one wishes to force an operation to be scalar, such as merely multiplying corresponding cells of matrices x and y, the dot modifier must be applied to the multiply operator as follows:

z = x .* y;

Octave functions are declared in the same way as MATLAB functions. It is also possible to have functions that receive a variable number of arguments. Octave uses the varargin array method of passing variable number of arguments to functions. For the user, this means that functions must have a bit of logic to default arguments, then check if values for them were passed in the varargin array.

One small difference between MATLAB and Octave is that comment lines in Octave can start with either the % character or the # character.

% This is a MATLAB or an Octave Comment
# This is also an Octave Comment

Linux Survey

In Linux, Octave Makes A Great Batch Language

In Linux, Octave is also easy to use as a batch language. One simply uses Octave like another shell. The following example would execute Octave as the shell and perform whatever Octave commands are in the file. In this little example, Octave prints out how many arguments (argv) are passed in. The arguments could easily be parsed to be used in the program:

#!/usr/bin/octave -qf
printf("You passed %d arguments\n",length(argv));

Using Octave in this way allows one to quickly script out solutions to problems, yet run them as one would run any other data processing program.

Octave Uses Simple and C-Style I/O

Octave only has built high level I/O routines for a few file types. Octave has simple commands for saving and loading matrices, and commands that import and export some of the older MATLAB compatible files. Octave only reads in it's own ASCII format image file format. It can output either its own format or pnm format files (ppm for color, pgm for gray scale).

However, Octave does provide C style I/O primitives for reading and writing all manner of user designed ASCII and binary files. So one could conceivably read or write any kind of file with Octave, as long as the format of the file is known. I was easily able to add a simple fits file reader to Octave for handling fits images, and another for reading the pnm ppm and pgm files.

Octave Can Do Many Types Of Graphs

octave contour plot

Octave, in Linux, uses a number of graphic programs in support. Most line graphs are done with Gnuplot. Many easy to use plot commands common to other languages, such as plot, contour, and mesh are made available, and use gnuplot for the presentation. The plot you see above is an example of a plain contour plot.

octave contour plot

The plot you see above is a more elaborate version of the kind made with the Octave mesh command. If you compare it with the plot in the introduction section, you'll see that it has both the 3D mesh of the image, and a contour graph on the x-y plane. There are also contours, color coded with those on the x-y plane, drawn through the 3D mesh.

Octave doesn't directly provide a command for a plot with all the features of the shown graphic. But Octave does provide additional primitive commands for controlling gnuplot, so it is possible to add more graphic commands (through m files) to get more performance from the very capable gnuplot program, such as this augmented mesh plot.

Octave can also display images, such as digital photographs like the one of the lunar crater Plato shown above. First, unfortunately, one has to convert the image to the Octave image format, but Octave can then load the image and display it. Rather than gnuplot, which is a line drawing graphic program, Octave passes the image to the Imagemagick display program (in Linux). Display then gives some additional control of the image, such as scaling, rotation, and other commands. Newer versions of octave can use other display utilities.

In working with Octave graphics commands, one thing I didn't find with the gnuplot or display integrations was a way to interact with graphic windows. That is, I was unable to find a way to receive mouse locations from within an Octave plot.

This can be rectified in Linux by alternatively using the plplot graphics package. Plplot has an Octave interface and can be used to replace the gnuplot graphics. With plplot installed, one can issue the toggle_plplot_use command and have access to another set of high level plot commands. With the plplot interface, mouse locations can be returned to an Octave program.

You can get help when in Octave by typing help -i. This will bring up a Linux info information file with a hierarchal help system. You can also type help routine, where routine is a documented function. Octave makes it easy for the user to document his or her own functions, as the documentation for each user generated function is in the function file.


Octave, with its high level of comparability with MATLAB, is a Linux favorite. It's matrix routines are quite fast, and with the addition of the Imagemagick and plplot supplemental graphics routines, provides a good platform for a wide range of mathematical problems. It's used heavily for signal processing and general statistical analysis. Because of the limitations of the graphics system and image file formats recognized, Octave isn't probably the first choice for image processing.

None-the-less, if you do as I did and add fits file and pnm file readers, you can work on images in octave. One note, if you decide to do this: octave works only with 2 dimensional matrices. Languages more suited to image processing typically hold color images in 3 dimensional matrices, with the 3 layers respectively holding each pixels red, green, and blue levels.

Octave uses a color map to supplement its 2 dimensional matrix format. With color images, what works is to create a color map that has as many rows as the number of pixels in an image, and 3 columns. The 3 columns are for the red, green, and blue color components. The image matrix than becomes just a matrix who's elements run from 1 to the number of cells in the image. Thus each cell just points to the respective row in the color map which has the 3 color levels for that image.

For example, assume a 4 x 4 pixel color image (I know that's lame, but it illustrates the point. In Yorick, the read in image would be stored in a matrix of 3 dimensions, like (3,4,4), for 3 layers of the 4 x 4 image: one for red, green, and blue color levels. Assuming 256 levels of color, that might look like:

Red Layer
209 193 179 157
178 116 189 187
66 86 133 133
69 57 51 59
Green Layer
198 178 161 137
157 85 147 148
57 64 95 90
51 39 34 40
Blue Layer
210 196 184 168
165 95 147 147
67 70 100 93
57 45 37 46

In octave, the image matrix and color map would look like this:

Image (Indexes Into Color map)
1 2 3 4
5 6 7 8
9 10 11 12
13 14 15 16
Color map * 255 (rgb per pixel)
209 198 210
193 178 196
179 161 184
157 137 168
178 157 165
116 85 95
189 147 147
187 148 147
66 57 67
86 64 70
133 95 100
133 90 93
69 51 57
57 39 45
51 34 37
59 40 46

Notice that the first row of the color map is respectively the first element of the red, green, and blue matrices. Then next row is the 2nd element of the first row of red, green and blue. Then see that the each element of the image matrix is simply the row number of the color map for that pixel.

One final note, the color map listed is multiplied by 255 (for 255 colors). In Octave, the actual color map values must be between 0 and 1. Dividing the illustrated color map by 255 would provide a usable octave color map for our little image example.

Below are listed my subjective pros and cons for the Octave Matrix language.

  • Freely available for Linux and Windows. The Linux version is easier to use IMHO.

  • Highly compatible (95%) with the commercial Matlab language, a terrific boon to college students.

  • Loaded with engineering problem solution utilities, such as signal processing.

  • An easy to learn language style.

  • Makes many C-Style file I/O utilities available, making it easy to write Octave routines to be compatible with files made with external utilities.

  • Has a reasonable number of string handling utilities, such as compare, search and replace, and split.

  • Very fast when solutions primarily use the intrinsic Matrix operators

  • Uses the handy m file system where each function file contains a single function, and simply referencing functions causes them to be auto-loaded.

  • Cons:
  • Works only with 1 and 2 dimensional matrices. As the above discussion of Octave colormaps illustrates, handling things like 3 color images is awkward.

  • Slows down considerably if solutions contain looping logic.

  • Primarily uses GNUPLOT for plotting. GNUPLOT is very versatile, but the combination doesn't provide the ability to have mouse actions in a plot return values to Octave.