GDB Armadillo helpers

This included pretty printers for armadillo vectors, matrices and cubes, as well as a few xmethods.

It also includes a pretty printer for std::complex<int> and std::complex<double> to make the output nicer.

Using these pretty printers

Clone this repository to some folder and add the code below to the .gdbinit in your home folder

source /path_where_you_cloned/gdb_armadillo_helpers/gdb_helpers/gdb_armadillo_printers.py
source /path_where_you_cloned/gdb_armadillo_helpers/gdb_helpers/gdb_std_complex_printer.py

After that just use p some_variable in gdb to see the result nicely formatted using gdb native format for arrays. This means that it will work better if you have set print array on in your .gdbinit file.

Note: This also works inside CLion and possible in other IDEs.

Screenshots

Consider the variables below

std::complex<int> c1(1, -3);
std::complex<double> c2(1.5, -3.3);
std::complex<double> c3(1.5, 3.3);

arma::mat m1{{1.1, 2.2, 3}, //
               {4, 5, 6},     //
               {7, 8, 9},     //
               {10, 11, 12},  //
               {13, 14, 15},  //
               {16, 17, 18}};

arma::cx_mat m2{{1.1 - 1.1j, 2.2 - 7.7j, 3},
                {4, 5, 6},
                {7, 8, 9},
                {10, 11, 12},
                {13, 14, 15},
                {16, 17, 18}};

When print the complex numbers without the pretty printers we get

With the pretty printers we get (notice the complex numbers are shown both in rectangular form and in polar form)

For the armadillo variables, without the pretty printers we get

You can see some internal data and the dimension of the matrix, but not the elements. It is possible to see the elements using gdb support for creating arrays and the fact that the stored elements are pointed by the mem variable. That is, use the command

p (*m1.mem)@6@3

While this works, it requires manually specifying the dimensions and it is not directly accessible. On the other hand, with pretty printers we get (notice that the dimensions is also shown and how arma::cx_mat uses the pretty printer for complex numbers)

This will even work in IDEs, such as CLion, as shown below

Note: The dimensions in the printing for matrices were swapped on purpose. That is, the first dimension indicates the column index, while the second one indicates the row index.

Configuration

You can set the value of the arma-show-content parameter to enable (default) or disable the armadillo pretty printers to show the elements. At any time, if you are only interested in vec/mat/cube dimension, the use set arma-show-content off in gdb and the armadillo pretty printers will only display the dimensions. Set the value to on to print the elements again.

Likewise, if you don't want the polar form of complex numbers to be shown you can run in gdb the command set complex-show-polar off.

Note: The pretty printers are also affected by gdb's native configuration for arrays, such as set print array on/off and set print elements SOME_NUMBER.

XMethods

XMethods are a feature of GDB python API that allow the re-implementation of C++ methods in Python in order for GDB to use. These C++ methods might not be available due to being inlined, optimized out, or simply because there is no inferior running (you are debugging from a core file, for instance).

The currently implemented xmethods are:

• min
• max
• size
• empty
• at (linear indexing for vectors, matrices and cubes, as well as 2D index for matrices and 3D indexing for cubes)
• slice (for cubes)

In order to have them available, add the code below to the .gdbinit in your home folder

source /path_where_you_cloned/gdb_armadillo_helpers/gdb_helpers/gdb_armadillo_xmethods.py

Custom 'print-numpy-array' gdb command and 'get_array' python function

Add the code below to the .gdbinit in your home folder

source /path_where_you_cloned/gdb_armadillo_helpers/gdb_helpers/gdb_armadillo_to_numpy.py

Now you can call the print-numpy-array gdb command passing the name(s) of any variables in the current scope which are armadillo types. Note that you can complete the variable names with TAB.

If you need more power, you can start the python interactive terminal from gdb with the the pi command (python-interactive). Once you are in the python terminal inside gdb, use the code below to get a numpy array from a variable called m.

m = get_array(gdb.parse_and_eval('m'))

# Just the name is also enough
m = get_array('m')

From there you can manipulate the numpy array as you want (print, compute the norm, etc). You can even run the python interactive terminal from gdb (the pi command), import matplotlib and plot the values in the matrix.

Note: Changes to this numpy array are not propagated in any way to the original m variable.