WeBWorK Problems

Ranking Tasks

Ranking Tasks

by Michael Gallis -
Number of replies: 4
I use WeBWorK (currently version 2.2.1) for my physics courses, and I was trying to develop a custom answer evaluator for ranking tasks. Students would rank items (such slopes at points on a graph, or numerical answers) . For example, if A=1, B=3, C=5 and D=3 students would input C>B=D>A or C>D=B>A (either acceptable). The ranking would be randomly generated for each student.

I spent a bit of time playing with Perl and decided that using hashes and built in hash/list tools would be the way to go. I developed some snippets that seemed to do the tasks I needed, but when I tried them within a trial problem, I got caught with this message:
'sort' trapped by operation mask at line 30 of (eval 1433)
which seemed to tell me that the sort operation in use
30 foreach $key (sort mrg_hashValueAscendingNum (keys(%rankingtask_scores))) {
31 $rankingtask_item_order{$key}=$order;
32 $order=$order+1;
33 }
was not allowed in .pg

Which (finally) brings me to my questions:
  • Am I reinventing the wheel? Are already question and evaluator styles and mechanisms that do what I am trying to do, but called something else? I didn't see any likely candidates in the wiki, cut could easily have missed.
  • Is it possible to get sort to work within a question and/or within an answer evaluation?
  • Is there any heavier duty tutorials on writing answer evaluators? I keep running into the old doc's from 1.6 and I'm not sure everything works the same.
-Mike Gallis

In reply to Michael Gallis

Re: Ranking Tasks

by Michael Gage -
sort itself is not allowed in .pg since it could cause
denial-of-service problems it was misused. There is a PGsort however which (supposedly) protects against this.

in PGbasicmacros.pl you have

 lex_sort(@list); # outputs list in lexigraphic (alphabetical) order
 num_sort(@list); # outputs list in numerical order
 uniq( @list); # outputs a list with no duplicates. Order is unspecified.
 PGsort( \&sort_subroutine, @list);
 # &sort_subroutine defines order. It's output must be 1 or 0 (true or false)
One of these will solve your immediate problem.

We're working on better tutorials (everyone welcome to help :-) ). The PG macro/library link in the wiki's Author's guide page is a very useful addition. There is also a database of almost all known PG macros at https://webwork.maa.org/moodle/mod/data/view.php?id=194

For the most part what you find in the 1.6 tutorials will still work in the current WeBWorK, but often there is a better way to do things -- particularly if you can put Davide Cervone's MathObjects to use.

In this particular case there might be some MathObjects (contextInequalities.pl) that would help evaluate simple equations. See the thread between Robin Cruz and Davide Cervone https://webwork.maa.org/moodle/mod/forum/discuss.php?d=5872

It's likely that using them will make the problem easier to write and to maintain.

I expect Davide will have more to say about this when he gets a chance.

-- Mike

In reply to Michael Gage

Re: Ranking Tasks

by Michael Gallis -
Thanks, those pointers are VERY useful. Unfortunately I'm still running into a sort of a sort problem.

The PGsort looks like it should be just what I need, but I seem to be having trouble getting it to work. Using the code snipet (adapted from the example at http://webhost.math.rochester.edu/webworkdocs/docs/pglanguage/manpages/pg_sort/)
in a problem:

$A = join(" ",PGsort( sub {$_[0] <=> $_[1]} ,(23,2,10,11,11,31) ) );
$A $BR

the result is
31 11 11 10 2 23

which is just the reverse order of the original list. I've gotten similar results from other experiments with PGsort, which at least suggests to me that PGsort isn't getting or using the sort routine correctly. Is it possible that my older version of WebWorK (2.2.1) has a buggy version of PGsort?

I am trying to use a hash to sort, along the lines of
PGsort( sub { $myhash{$_[0]} <=> $myhash{$_[1]}} , @mylist )
which I've gotten to work in perl test scripts using perl's native sort.

Thanks for any directions you can provide.

-Mike Gallis
In reply to Michael Gallis

Re: Ranking Tasks

by Davide Cervone -
The problem is that PGsort doesn't expect the subroutine to return the result of the <=> operator, but rather just 1 or 0 depending on whether the first is less than the second or not. This is documented in
near the bottom (look for "Sorting and other list macros"). I'm not sure why they decided to do it that way, but that is the way it works.


In reply to Michael Gage

Re: Ranking Tasks

by Davide Cervone -
I don't think the Inequalities context will help with this, as it is just an alternative syntax for obtaining Interval objects, so would not work with the kind of notation for sorted lists that is being requested.

I think it would probably require the development of a specialized context for this. It shouldn't be too hard to do, but I'm not sure when I'm going to bale to take time for it. I'm also not sure of all the parameters that are needed. For example, will the terms always be letters, or will it include numbers? Will there ever be operations for the terms (like 10 > 2+6 > 4)? What should we say about errors like 10 > 7 < 8? Should there only be one of the two inequalities? Should there be >= and <=?