Hi Rob!
On Tue, Dec 14, 2010 at 03:09:55PM -0800, Rob Beezer wrote:
> Symptom: The following code would appear to run forever, since the
> rationals define an iterator that never quits.
>
> QQ.list()
>
> I discovered this when I presumed that
>
> (QQ^2).list()
>
> would raise an error. It doesn't. So I patched the free module code
> so the iterator producing module elements would raise an error for a
> module over an infinite base ring. That very nearly passed all tests,
> except it ran afoul of the category test suite. A doctest for sets
> (in the Set_object_union class) constructs a vector space. To get a
> generic element for testing the test suite would appear to build an
> iterator and then grab the first element produced, so my fix broke
> that. (This appears to be at
> categories.enumerated_sets._an_element_from_iterator)
>
> It looks like structure.parent._list_from_iterator_cached is where
> the first examples get stuck, at least that looks like the right place
> when you read the code and that's part of the traceback when you use
> ctrl-C.
>
> Options:
>
> 1. _list_from_iterator_cached could ask if self is infinite and if
> so, fail. This would break the test suite too, I think.
>
> 2. The test suite framework needs to get an element by some other
> process. For example, (QQ^2).an_element() did not break with my
> patch.
>
> 3. Tough. You ought to check if a set is infinite before you try to
> list it (and students should just learn to do that).
>
> 4. Something else.
>
> I've attached one failure of the test suite from my patch to the
> module code, there were four very similar failures from the one test:
> _test_an_element, _test_elements, _test_elements_eq,
> _test_some_elements. The last error message is from my patch.
>
> I'd started this as: http://trac.sagemath.org/sage_trac/ticket/10470
>
> I would prefer to have modules raise an error rather than just hang.
> If desired, I can try to fix more fundamental situations, like
> ZZ.list(). Comments and suggestions appreciated.
The list method from InfiniteEnumeratedSets() does just that. So here
is a possible route:
- There should be an InfiniteSets() category
- QQ should be in this category
- The list method from InfiniteEnumeratedSets() should be lifted to
InfiniteSets()
Shorter alternative: QQ could be in InfiniteEnumeratedSets(). But for
which one of the possible enumerations?
For the record, I am planning to work in the coming month(s?) on
improving the category framework to support "variants" (see [1]); this
should include InfiniteSets as a side product.
Cheers,
Nicolas
[1] http://trac.sagemath.org/sage_trac/wiki/CategoriesRoadMap
--
Nicolas M. Thiéry "Isil" <nthi...@users.sf.net>
http://Nicolas.Thiery.name/
--
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org
