Improve question-based learning progress calculation
The issue was that "question-based" learning progress was relying on the underlying score system that we implemented with the very first version of the learning progress feature. The idea of using points stems from the usual setting in academia where points are awarded for answering questions. If only parts of a question are correct, the learner would still get credit for those parts. This is the score or point-based learning progress calculation.
However, it turned out that we sometimes just want to know if a question has been answered correctly or not, and we are not concerned about partially correct answers. Hence, we needed a "question-based" learning progress. To calculate this learning progress, we used the underlying scoring system. That system made it pretty easy to implement new calculations on top of existing answers. It also allowed us to use very small database views.
As it turned out, there are some rare cases where points are not correctly applied to questions. We encountered this problem in a recent lecture where participants provided a list of questions they had prepared in their own sessions, then exported them and sent them to the lecturer. The lecturer imported all questions from all participants into a master ARSnova session. Participants could then answer all these combined questions to receive a grade based on the learning progress' readings.
Unfortunately, question export/import does not carry over the points that were initially set on a question. We were now in a situation were people had answered questions and were supposed to be graded on them, but couldn't get any learning progress -- since there were no points awarded for any question.
Letting everyone re-answer all questions after fixing the assigned points for every question was of course not possible -- we wanted to grade the students after all -- so we needed a different solution.
We decided that "question-based" would mean using the questions itself to grade the user, and not points derived from those. There is also a usability issue here as the naming does not suggest the usage of points because the correct answers can be marked independently.
That is how we arrived at the current implementation.
Note: The formulas did not change, just the means to generate the data to feed into the formulas has changed.