Question

This is kind of convoluted, so if I’m missing out on an easy construct for this, please let me know 🙂

I’m analysing the results of some matching experiments. At the end game, I want to be able to query things such as experiments[0]["cat"]["cat"], which yields the number of times “cat” was matched against “cat”. Conversely, I could do experiments[0]["cat"]["dog"], when the first query was a cat and the match attempt was a dog.

The following is my code to populate this structure:

    # initializing the first layer, a list of dictionaries.
    experiments = []
    for assignment in assignments:
        match_sums = {}
        experiments.append(match_sums)


for i in xrange(len(classes)):
        for experiment in xrange(len(experiments)):
            # experiments[experiment][classes[i]] should hold a dictionary,
            # where the keys are the things that were matched against classes[i], 
            # and the value is the number of times this occurred.
            experiments[experiment][classes[i]] = collections.defaultdict(dict)

            # matches[experiment][i] is an integer for what the i'th match was in an experiment.
            # classes[j] for some integer j is the string name of the i'th match. could be "dog" or "cat".
            experiments[experiment][classes[i]][classes[matches[experiment][i]]] += 1
            total_class_sums[classes[i]] = total_class_sums.get(classes[i], 0) + 1

    print experiments[0]["cat"]["cat"]
    exit()

So clearly this is a bit convoluted. And I’m getting a value of “1” for the last match, rather than a full dictionary at experiments[0]["cat"]. Have I approached this wrong? What could the bug here be? Sorry for the craziness and thanks for any possible help!

Answer 1

Two points:

• Dictionary keys can be tuples; and
• If you’re counting things, use collections.Counter. (You can use defaultdict(int), but Counter is more useful.)

So, instead of

experiments[experiment][classes[i]][classes[matches[experiment][i]]] += 1

write

experiments = Counter()
...
experiments[experiment, classes[i], classes[matches[experiment][i]]] += 1