Powers with rational exponents from QQbar fail

[rws]

sage: 2^(QQbar(1))
...
TypeError: no canonical coercion from Algebraic Field to Rational Field

[rws]

New commits:

​d3f8911

21740: catch cases that can be converted to rational

[tscrim]

The e = QQ(e) should call _coerce_ first then call the conversion, so _coerce_ shouldn't need to be called again. (We'd also then want e = QQ(e) to raise the error.)

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​bcdf0ed

21740: change order of calls

[rws]

You mean, this way?

[nbruin]

If you're going to try conversion anyway, I don't think it makes much sense to try coercion explicitly first. I think it's good custom that if a coercion exists, conversion should agree with it. So, in effect (and possibly even in implementation!), conversion would try a coercion already.

One of the reasons why conversion might not be appropriate here is that, for instance, QQ("2") works just fine. So, if you use conversion in exponents the following probably would hold:

sage: 2^"2"
4
sage: a=GF(3)(-1)
sage: 2^a
4

That's a little too lenient to my taste.

[jdemeyer]

As me and Vincent Delecroix wrote on sage-devel, it is a feature that 2 ^ QQbar(1) raises an exception.