COMP0233: Research Software Engineering With Python


Using a debugger

Stepping through the code

Debuggers are programs that can be used to test other programs. They allow programmers to suspend execution of the target program and inspect variables at that point.

Using the python debugger

Unfortunately this doesn't work nicely in the notebook. But from the command line, you can run a python program with:

python -m pdb my_program.py

Basic navigation:

Basic command to navigate the code and the python debugger:

  • help: prints the help
  • help n: prints help about command n
  • n(ext): executes one line of code. Executes and steps over functions.
  • s(tep): step into current function in line of code
  • l(ist): list program around current position
  • w(where): prints current stack (where we are in code)
  • [enter]: repeats last command
  • anypythonvariable: print the value of that variable

The python debugger is a python shell: it can print and compute values, and even change the values of the variables at that point in the program.


Break points tell debugger where and when to stop We say

  • b somefunctionname
In [1]:
%%writefile solutions/diffusionmodel/energy_example.py
from diffusion_model import energy
print(energy([5, 6, 7, 8, 0, 1]))
Writing solutions/diffusionmodel/energy_example.py

The debugger is, of course, most used interactively, but here I'm showing a prewritten debugger script:

In [2]:
%%writefile commands
restart  # restart session
b energy # program will stop when entering energy
c        # continue program until break point is reached
print(density) # We are now "inside" the energy function and can print any variable.
Writing commands
In [3]:
python -m pdb solutions/diffusionmodel/energy_example.py < commands
> /home/runner/work/rsd-engineeringcourse/rsd-engineeringcourse/ch03tests/solutions/diffusionmodel/energy_example.py(1)<module>()
-> from diffusion_model import energy
(Pdb) Restarting /home/runner/work/rsd-engineeringcourse/rsd-engineeringcourse/ch03tests/solutions/diffusionmodel/energy_example.py with arguments:
	# restart session
> /home/runner/work/rsd-engineeringcourse/rsd-engineeringcourse/ch03tests/solutions/diffusionmodel/energy_example.py(1)<module>()
-> from diffusion_model import energy
(Pdb) > /home/runner/work/rsd-engineeringcourse/rsd-engineeringcourse/ch03tests/solutions/diffusionmodel/energy_example.py(2)<module>()
-> print(energy([5, 6, 7, 8, 0, 1]))
(Pdb) Breakpoint 1 at /home/runner/work/rsd-engineeringcourse/rsd-engineeringcourse/ch03tests/solutions/diffusionmodel/diffusion_model.py:2
(Pdb) > /home/runner/work/rsd-engineeringcourse/rsd-engineeringcourse/ch03tests/solutions/diffusionmodel/diffusion_model.py(9)energy()
-> from numpy import array, any, sum
(Pdb) [5, 6, 7, 8, 0, 1]

Alternatively, break-points can be set on files: b file.py:20 will stop on line 20 of file.py.


Debugging when something goes wrong:

  1. Have a crash somewhere in the code
  2. run python -m pdb file.py or run the cell with %pdb on

The program should stop where the exception was raised

  1. use w and l for position in code and in call stack
  2. use up and down to navigate up and down the call stack
  3. inspect variables along the way to understand failure

This does work in the notebook.

%pdb on
from diffusion.model import energy