Break learning up into small task-based tests for focused study.

Tests can be defined with prerequites so a student wishing to learn a higher level task, but out of their depth with the task itself, can easily go back through the basics. By default, only the leaf questions (i.e. questions that are not dependencies of other question) are asked. If the user gets one wrong, we push the question back on the stack (so they can try again later), and also push all of that questions direct dependencies onto the stack (so they can get the background they need to answer the question they got wrong).

There are a number of example quizzes available in the quizzes directory. The example quizzes mostly focus on teaching software development tasks (POSIX shell utilities, Git version control, …), but any material that can be presented in a textual prompt/response/check process should be fairly easy to develop. The quiz files are written in JSON, and the format should be fairly easy to understand after looking through the examples.

The quiz framework and answer processing are independent of the user interface used to present the prompts and collect responses. Currently only a input() based command line interface exists, but other interfaces (e.g. a web server for browser-based interaction) should be fairly straightforward.

Here's an example typescript for one of the sample quizzes:

$ ./pq.py quizzes/monty-python.json
What is your favourite color?
? blue
correct

What is the capital of Assyria?
? Hmm…
incorrect

What is your quest?
? To seek the Holy Grail
correct

What is the capital of Assyria?
? ?

What is the capital of Assyria?
Sir Robin didn't know it either
? I don't know that
correct

results:
question:     What is your quest?
answers: 1/1 (1.00)
  you answered: To seek the Holy Grail
     which was: correct

question:     What is your favourite color?
answers: 1/1 (1.00)
  you answered: blue
     which was: correct

question:     What is the capital of Assyria?
answers: 1/2 (0.50)
  you answered: Hmm…
     which was: incorrect
  you answered: I don't know that
     which was: correct

answered 3 of 4 questions
of the answered questions, 3 (1.00) were answered correctly

The unanswered question (“What is your name?”) wasn't asked because the user successfully answered the question that depended on it (“What is your quest?”).

Quizzer requires Python ≥ 3.2. If Pygments is installed, the command line prompt will be colored.

Types of questions:

• Question: An open ended question with a single correct answer (or accepting any answer).
• NormalizedStringQuestion: Like Question, but normalizes the given answer with string.strip().lower() before comparing with the correct answer.
• ChoiceQuestion: A question with multiple valid answers. With display_choices, this is a standard multiple choice question. With multiple_answers, it is “chose all that apply”.
• ScriptQuestion: A question which checks scripting answers by running the given answer and correct answer in temporary directories. There are hooks for setting up and tearing down the temporary directories, and we judge success by comparing the output of the teardown phase (and optionally the output of the answer phase). You can optionally set a timeout to catch answers that hang, although this requires Python ≥ 3.3. There is no sandboxing (other than working in a scratch directory), so it's not a good idea to serve questions like this on a public interface (stick to cli).