Rewritten my problem-tester project from scratch in Python 3. Here are the improvements.
Debugging
The problem with the old solution was that it ran shell-exec to generate the output. That meant debugging wasn’t possible.
Now you can debug your scripts with PyCharm thanks to a feature of Python 3.
With Python 3 it is possible to redirect STDIN and STOUT by doing
fh_in = open(f_in, 'r') fh_out = open(f_out, 'w') sys.stdin = fh_in sys.stdout = fh_out
We now need to include the problem file and run it in a loop, passing “.in” and “out” files. Which wasn’t that simple.
The first challenge was to import a python module dynamically, not knowing its name in advance. Thankfully there is an answer on the SO regarding that, I just needed to adjust it a little.
The next problem was to reload the module and run it freshly for every new input. The best solution I found was to delete it from sys.modules and import again. I’m yet to check if it has any side-effects but so far it looks good.
def run_problem(mod):
# remove module
if mod in sys.modules:
del sys.modules[mod]
# dynamically import it again.
problem = __import__(mod, globals(), locals(), ['main'], 0)
# if the module contains 'def main' execute it.
# If not, we assume the script was in global scope, importing it forced a run
# Nothing to do
if 'main' in dir(problem):
problem.main()
Validating
Sometimes there may be several solutions to a problem, like here: http://codeforces.com/problemset/problem/873/D. In this case you can write a validator.py script that tests the solution.
Comparing with another solution
If you have a solution that works, you can compare put it with the name solver.py. It will generate “ok.out” files.