I established the following requirements together with Gustav:
1. A quiz can be attempted multiple times, where each instance of a quiz may be differently randomized.
2. While attempting a quiz, individual questions may be attempted multiple times. In this case the same instance of a question should be used always.
Implicit requirements from the current quiz module are:
3. The number of attempts per quiz may be restricted.
4. The number of attempts per question (in a particular quiz) may be restricted.
Further possible requirements:
5. It may be sensible to allow students to review or continue working on a quiz at a later stage.
To get a better understanding I drew a diagram of the current DB structure (omitting tables that I do not consider relavant here). Please correct me if I'm wrong here!
I agree with Henrik that no changes to the database are required to achieve this, however, it seems sensible to clean up the model somewhat,
a) to make it easier to understand,
b) to make it more efficient and
c) to make it (more) compliant with relational database theory.
Therefore I would suggest to eliminate the direct connection between 'quiz' and 'quiz_questions', which is currently implemented as a comma-separated list in the 'quiz' table. This violates the
1st normal form of relational database theory. Since the 'quiz_question_grades' table already exists, which essentially provides the same information, I suggest to eliminate the 'questions' field in the 'quiz' table and to store any information, that is specific to a question in a particular quiz (i.e. the grade) in the 'quiz_question_grades' table (and maybe rename it to 'quiz_question_information').
The 'quiz_responses' (each attempt on a question, a better name may be 'quiz_question_attempt') should then be linked ith the 'quiz_question_information table, because the response is given to a particular question in a particular quiz. There does not seem to be any immediate benefit to this, however it seems more consistent to me.
Given randomised questions (this is of particular interest to us) in a 'quiz_attempt' it becomes necessary to provide a means to reconstruct the exact version of a question a student has worked on (e.g. a random-seed). If a random-seed were stored in the 'quiz_attempts' table, a question-instance could be reconstructed for each 'quiz_response'. (The default questiontype would probably need to be extended to support randomisation given a random seed).
Finally, the 'quiz_grades' table seems to duplicate the 'quiz_attempts' table to some extent. I assume that it stores the grade that is currently assigned to a student (probably the grade that was displayed to the student). I am not entirely sure if it is necessary to store this information, since it can be quite easily derived by looking at all 'quiz_attempts' of a user on a particular 'quiz' and applying the chosen grading algorithm.
The proposed changes would result in a new DB structure as shown in the diagram below.
Please comment on this!
Julian
