#14793: Unique representation for homsets
-------------------------------+--------------------------------------------
Reporter: nthiery | Owner: nthiery
Type: enhancement | Status: needs_review
Priority: major | Milestone: sage-5.12
Component: categories | Resolution:
Keywords: | Work issues:
Report Upstream: N/A | Reviewers:
Authors: Simon King | Merged in:
Dependencies: | Stopgaps:
-------------------------------+--------------------------------------------
Changes (by SimonKing):
* status: new => needs_review
Old description:
> The unique representation of homsets is taken care of by ``Hom``. What's
> missing is:
>
> - Fast hash and comparison by id. This can be implemented by having
> Homset inherit from WithEqualityById
> - Pickling by construction, calling back Hom(domain, codomain, category)
New description:
The unique representation of homsets is taken care of by ``Hom``. What's
missing is:
- Fast hash and comparison by id. This can be implemented by having Homset
inherit from WithEqualityById
- Pickling by construction, calling back Hom(domain, codomain, category)
__Apply__
* [attachment:trac14793-homset_default_pickling.patch]
--
Comment:
I think I have solved all remaining problems. I needed to modify the
suggested logic of "relaxing the category test when there is reason to
believe that Hom is called on a domain that is in the process of being
unpickled".
It now reads like this:
{{{
#!python
# Determines the category
if category is None:
category = X.category()._meet_(Y.category())
# Recurse to make sure that Hom(X, Y) and Hom(X, Y, category) are
identical
H = Hom(X, Y, category)
else:
if not isinstance(category, Category):
raise TypeError, "Argument category (= %s) must be a
category."%category
# See trac #14793: It can happen, that Hom(X,X) is called during
# unpickling of an instance X of a Python class at a time when
# X.__dict__ is empty. In some of these cases, X.category() would
# raise a error or would return a too large category (Sets(), for
# example) and (worse!) would assign this larger category to the
# X._category cdef attribute, so that it would subsequently seem
that
# X's category was properly initialised.
# However, if the above scenario happens, then *before* calling
# X.category(), X._is_category_initialised() will correctly say
that
# it is not initialised. Moreover, since X.__class__ is a Python
# class, we will find that `isinstance(X, category.parent_class)`.
If
# this is the case, then we trust that we indeed are in the
process of
# unpickling X. Hence, we will trust that `category` has the
correct
# value, and we will thus skip the test whether `X in category`.
try:
unpickle_X = (not X._is_category_initialized()) and
isinstance(X,category.parent_class)
except AttributeError: # this happens for simplicial complexes
unpickle_X = False
try:
unpickle_Y = (not Y._is_category_initialized()) and
isinstance(Y,category.parent_class)
except AttributeError:
unpickle_Y = False
if unpickle_X:
cat_X = category
else:
try:
cat_X = X.category()
except BaseException:
raise TypeError, "%s is not in %s"%(X, category)
if unpickle_Y:
cat_Y = category
else:
try:
cat_Y = Y.category()
except BaseException:
raise TypeError, "%s is not in %s"%(Y, category)
}}}
Note the comment on simplicial complexes: They do have a `.category()`
method, but they do not derive from `CategoryObject` and thus have no
`._is_category_initialized()` method. Shame on them.
In addition, some changes needed to be done for `SchemeHomsets` and the
like. I think one can now start with reviewing.
Apply trac14793-homset_default_pickling.patch
--
Ticket URL: <http://trac.sagemath.org/sage_trac/ticket/14793#comment:31>
Sage <http://www.sagemath.org>
Sage: Creating a Viable Open Source Alternative to Magma, Maple, Mathematica,
and MATLAB
--
You received this message because you are subscribed to the Google Groups
"sage-trac" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/sage-trac.
For more options, visit https://groups.google.com/groups/opt_out.
