Hi Volker,
> I agree that I need to set the attribute Element not element_class in my
> path. I'll fix it after the dust in #10496 has settled.
>
> I'm still confused about _element_class(). If you need to compute the
> correct value of Element, then just do that in the constructor. So if
> _element_class() is only called in the constructor then I'm perfectly happy
> with that. But none of the derived classes of FGP_Module actually does that.
> And before I rewrote things, it was full of "return
> self.element_class()(V,W,check)" whenever an element was to be constructed.
>
> So my understanding is that parents should:
>
> * directly define myparent.Element if its always the same class
> * define myparent.Element in myparent.__init__() if it depends on the
> parent.
>
> Is that correct?
Just to make it clear:
element_class is obtained by sub-classing Element adding all the class
provided by the category system at the top of the mro. In principle, you
should not touch it unless you want to reuse a already computed element class
from a different parent in the same category.
you can define myparent.Element in any way that suits you including
1 - as a class attribute:
class myParent(...):
...
class Element(...):
...
or
class myParent(...):
...
Element = SomeClass
2 - as a lazy attribute
class myParent(...):
...
@lazy_attribute()
def Element(self, ...):
class Toto(...)
...
return Toto
3 - as a standard attribute:
class myParent(...):
def __init__(self, ...):
self.Element = ...
Parent.__init__(self, ...)
But in any way (and in particular for 3-) you have to define it *before*
calling the initialization of category in Parent.__init__.
Hope this helps,
Cheers,
Florent
--
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
