New question type request: numfunc n-tuple

[fzql]

Not sure what the proper procedure for feature request is, but I request a new type of question, the numfunc n-tuple. which is like an n-tuple or calculated n-tuple, but with numfunc entries.

Having coded questions whose answers are n-tuples of numfuncs, I see that this type of question could be very helpful in multivariable calculus. Example questions are gradients of functions, e.g., f(x,y,z)=xyz, where students have to respond with a vector of 3 numfuncs, each being yz, zx, and x*y. The current way in which I am coding this type of question is to use conditional questions together with some answer-parsing code can split student responses enclosed in single or double brackets (round or angle) into an array of strings.

If there is developer documentation, I would like to try to submit some commits. I am a bit short on time, so here is only my feature request.

[drlippman]

Is there a reason that this approach would be significantly preferable over a multipart question where they input each component separately?

[fzql]

By analogy, the current n-tuple and calculated n-tuple are all arguably preferable their multipart counterparts.

From another perspective, I inspected the current atom question types and they fall into one of the following categories:

• Question type with n-tuple support: number (ntuple), calculated (calcntuple),
• Question type without n-tuple support at the moment: choices, multans, matching, numfunc, draw, matrix, calcmatrix, complex, calccomplex, interval, calcinterval, essay, file, string

In this second category, numfunc is best suited to receive n-tuple support. I am not sure if complex and calccomplex should receive n-tuple support, but the rest in the second category probably do not benefit from n-tuple support at all and are best served by multipart.

However, I fail to think of any pedagogical reasons why numfunc n-tuples would benefit students other than that a potential interface for question input looks nicer and more in-line with calculated n-tuples and numfunc. In the course that I am teaching now, the students seem to be equally fine with the multipart and the conditional versions.

Edit: On the level of question design, official support for numfunc n-tuple would make the grading of parametric equations and vector fields more efficient and consistent across the question pool. Many options ($variable, $domain, $displayformat, $scoremethod) available to numfunc and calculated could also be made available to numfunc n-tuples automatically. But, of course, this question type has limited use outside of scope of vector-valued functions.