Ideal with no generators for univariate polynomial rings

[mraum]

Consider the following (Sage 4.7alpha1) :

sage: P.<a> = QQ[]
sage: P.ideal([])
Booom!

The issue is twofold: The ideal method in rings/rings.py is called and does not accept an empty list. But also the univariate polynomial rings should call a specialized method just like the multivariate ones do.

It worked in earlier versions (like 4.2 or so) and is a construction that easily comes up as a corner case.

Apply 11139_alternative.patch​.

[jhpalmieri]

I'm happy with this. Here's a new patch with a few doctests illustrating that the problem has been fixed.

I'm going to mark this as "needs review"; it wouldn't hurt to have someone else look it over.

[jdemeyer]

What do you think of this alternative patch? IMHO, it has better logic.

[mraum]

I think that indeed the logic is much better. Though it might be confusing to beginners. But this improvement brings my attention to a problem, that has been there before:

R.<x> = QQ[]
R.ideal((x for _ in range(2)))

This is an instance of GeneratorType?, that is called with len. So either we exclude the GeneratorType? (no change to the situtation before) or correct it. Also, mind that a ring element might be an instance of GeneratorType?. In that (very exceptional) situation, the new logic would cause trouble.

Any experience with similar code at other places in Sage?

[jdemeyer]

Replying to mraum:

This is an instance of GeneratorType?, that is called with len. So either we exclude the GeneratorType? (no change to the situtation before) or correct it.

I would say not to check for GeneratorType. I have added a new patch for this. I still have to test it thoroughly.

[jdemeyer]

Alternative patch, apply only this

[jdemeyer]

Patch tested, all tests passed.

[jdemeyer]

Any reviewers?

[mraum]

I think, since I haven't written this ultimate patch, I am a legitimate reviewer. All tests pass and also the patch bot does not complain. Although we have changed the logic, backward compatibility is preserved.