Make a good use of the EmptySetError in EnumeratedSets and Parent

[nborie]

As Florent Hivert gave extremely useful features about empty things, we should use it. The Following should returns an EmptySetError?

sage: S = Set([])
sage: S.an_element()
---------------------------------------------------------------------------
StopIteration                             Traceback (most recent call last)

...


StopIteration: 
sage: 

Also, make the EmptySetError? being documented already in the structure of Parent. Improve the documentation of an_element and _an_element_ in Parent.

[robertwb]

Technically, some_elements should return an iterable, not an iterator, so you can iterate over it more than once. The except clause should be changed to return [], not return iter([]), though with the empty case the behavior is the same.

[nborie]

Good point. I will ask today Nicolas what does he think about that... I don't really know the specifications for the Test Suite and how deal exactly with such corner case. Florent did this with FiniteEnumeratedSet? (The user know a priori that the set is finite...)

sage: FiniteEnumeratedSet([])
{}
sage: FiniteEnumeratedSet([]).some_elements()
<generator object _some_elements_from_iterator at 0xb433464>

I will give an answer shortly for that.

[nborie]

As this ticket fix a corner case and try to increase of the coherence for the code about sets, it have to verify all the specifications. Thanks you very much Robert for pointing this. It now return []. Even the behavior is the same, I don't want to fix a mathematical corner case by introducing a Spec corner case...

It should be ready for review now.

[zimmerma]

I confirm the problem is fixed, and all doctests still pass. Thus I give a positive review.

Paul

[jdemeyer]

I think you should add an example which shows the EmptySetError.

[nborie]

I just had two tests... Especially one which returns the EmptySetError?... Thanks all for your comments and suggestions.

[zimmerma]

after thought, the on-line documentation of S.an_element should mention that if S is empty, an EmptySetError is returned.

Paul

[robertwb]

Looking good now.

[jdemeyer]

You should change the commit message of the patch (using hg qrefresh -e) to something sensible, making sure the ticket number appears on the first line of the commit message.

[nborie]

In order to fix more things in the same time, I update the description of this ticket. I folded the NT'reviewer from the sage-combinat queue, then I uploaded the patch with a better commit message.

I am very OK with the changes in the Parent structure from Nicolas Thiéry. I give a positive review on all changes touching the corresponding file. The first part of the patch still comes from me, so no review from me.

[nthiery]

And positive review on your part. Good to go!