[LON-CAPA-dev] PGRD issues summary

[Michael T Hamlin]

This is a summary of a situation we are in and provides an overview
of some bugs/enhancements that would help us a great deal.  I've
entered a couple of them into bugzilla already, but this overview
perhaps makes it clearer what we're up against and how the bugzilla
entries hang together.

we have used several problems that have restrictions on significant
figures.  This has caused problems because most of our students don't
understand significant figures, and it isn't a topic we cover in our
course.

What I would like to happen, if it were possible, is that we turn off
significant figures globally (with appropriate course-wide parameter),
and then automatically regrade all submissions of the students who
currently have the problem incorrect.

This is a sort of multi-part challenge.  The first is selecting only
those students who have the problem incorrect, currently.  There is
(AFAIK) no easy way to do this; I have entered a bug #2102 for that.

The second is having responses automatically regraded, in sequence,
stopping if any of the responses is judged correct.  I see the problem
with trying this is that the _correct_ answer might have changed at
some point if the problem was republished with any changes that would
have changed the random seed, and judging an answer given days ago with
the _current_ correct answer, may give Unexpected Results.  In a perfect
LON-CAPA, because ALL versions of a problem are stored, the recorded
responses could be compared with the correct answers of the version of
the problem at the time the response was given.  I would like to suggest
this as a future enhancement that would save a lot of headache.  I have
not yet entered this in bugzilla (the other enhancements are necessary
first, I think).

In the shorter term, it would speed things up if we could PGRD the
problems in question, looking at only the current correct answer and
the set of responses entered.  Unfortunately the way PGRD is currently
set up, this procedure is much more time-consuming than it could be.
While it is possible, you have to look at a HUGE amount of information
that you don't need to see.
1.  The correct numerical answer is not shown unless you view the
ENTIRE problem text (bug #2236)
2.  You cannot select a particular problem part to grade, only the
whole problem.
3.  The responses of multipart problem are sorted chronologically,
instead of by part.
... and of course, the above issue with selecting the students you
need to regrade.

In the meantime, any suggestions or workarounds to help our situation
would be greatly appreciated.

michael