Better metaclass inference in dynamic classes

[saraedum]

When creating a dynamic class, (e.g., when creating an object in a category,) the following code is used to figure out the metaclass:

metaclass = DynamicMetaclass
# The metaclass of a class must derive from the metaclasses of its
# bases. The following handles the case where one of the base
# classes is a known Sage metaclass.  This approach won't scale
# well if we start using metaclasses seriously in Sage.
for base in bases:
    if isinstance(base, InheritComparisonClasscallMetaclass):
        metaclass = DynamicInheritComparisonClasscallMetaclass
    elif isinstance(base, ClasscallMetaclass):
        metaclass = DynamicClasscallMetaclass
    elif isinstance(base, InheritComparisonMetaclass):
metaclass = DynamicInheritComparisonMetaclass

This does not work if two bases have conflicting metaclasses, say InheritComparisonMetaclass and ClasscallMetaclass. As a result, one can currently not let a subclass of Morphism (which is a InheritComparisonMetaclass) inherit from UniqueRepresentation (a ClasscallMetaclass).

[saraedum]

I fixed a real world problem that hit me when working on #21879 to showcase that this really works. I could also split this off into a different ticket and create a more artificial doctest for this, if the reviewer wants me to.

---

New commits:

​e40f7a1	Better inference of metaclass when creating dynamic classes
​4c8be59	Implement a morphism that inherits from UniqueRepresentation
​6d4a953	fix typo
​726a9c3	Do not skip test
​4210389	Verify that unique representation works
​c84f7ad	Normalize arguments of AlgebraMorphism

[saraedum]

nthiery: I believe that you wrote the code that I quote in the summary?

[jdemeyer]

Replying to saraedum:

I fixed a real world problem that hit me when working on #21879 to showcase that this really works. I could also split this off into a different ticket and create a more artificial doctest for this, if the reviewer wants me to.

Please add such a doctest!

[jdemeyer]

And I would also recommend to split off the changes which are unrelated to the topic of this ticket (the changes to AlgebraMorphism for example).

[git]

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

​0f93b46

Add a doctest for 21895

[saraedum]

New commits:

​0f93b46

Add a doctest for 21895

---

New commits:

​0f93b46

Add a doctest for 21895

[jdemeyer]

1. Compare classes with is instead of ==.

2. Move the changes to generic_basis_code.py to a new ticket.

[git]

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

​a4e567f	use "is" to compare types
​86b4868	remove algebra morphism changes

[saraedum]

New commits:

​a4e567f	use "is" to compare types
​86b4868	remove algebra morphism changes

---

New commits:

​a4e567f	use "is" to compare types
​86b4868	remove algebra morphism changes

[saraedum]

I dropped the other change since I do not care for that particular case.