UniqueRepresentation and hash value

[hivert]

The documentation of UniqueRepresentation says

    Similarly, the identity is used as hash function, which is also as
    fast as it can get. However this means that the hash function may
    change in between Sage sessions, or even within the same Sage
    session (see below). Subclasses should overload :meth:`__hash__`
    if this could be a problem.

But there is no implementation of __hash__.

I'm adding one.

CC: @zimmermann6 @nthiery

Component: algebra

Keywords: UniqueRepresentation hash

Author: Florent Hivert

Reviewer: Paul Zimmermann

Merged: sage-4.3.3.alpha0

Issue created by migration from https://trac.sagemath.org/ticket/8120

[hivert]

Changed keywords from none to UniqueRepresentation hash

[hivert]

comment:2

I just submitted a patch which comply with the documentation. However, since we are in UniqueRepresentations we could use __classcall__ which knows the parameters that constructed the object to insert into the created object the hash value of those parameters in the same veins it insert some stuff needed for pickling/unpickling. Any comment about that ?

Florent

[hivert]

Author: Florent Hivert

[zimmermann6]

comment:3

I'd like to review that patch but I'm missing an example to try.

Paul

[hivert]

comment:5

Hi Paul,

Replying to @zimmermann6:

I'd like to review that patch but I'm missing an example to try.

I'm not sure what you want: in the patch you have the following tests example:

sage: class bla(UniqueRepresentation, SageObject): pass 
sage: x = bla(); hx = hash(x) 
sage: x.rename("toto")    
sage: hx == hash(x) 
True 

If you want more elaborated examples using UniqueRepresentation, they are dozens of them through sage library (the ultimate goal is that nearly each parent inherits from this). Pick you favorite one (prime.py is a good example see #7397):

tomahawk-*ge-combinat/sage $ grep -l UniqueRepresentation **/*.py
categories/category.py
categories/enumerated_sets.py
categories/examples/commutative_additive_monoids.py
categories/examples/commutative_additive_semigroups.py
categories/examples/finite_coxeter_groups.py
categories/examples/finite_enumerated_sets.py
categories/examples/finite_monoids.py
categories/examples/finite_semigroups.py
categories/examples/finite_weyl_groups.py
categories/examples/infinite_enumerated_sets.py
categories/examples/monoids.py
categories/examples/semigroups.py
categories/examples/sets_cat.py
categories/primer.py
categories/semigroups.py
categories/sets_cat.py
combinat/crystals/affine.py
combinat/crystals/letters.py
combinat/free_module.py
combinat/root_system/cartan_type.py
combinat/root_system/root_system.py
combinat/root_system/type_dual.py
combinat/root_system/type_relabel.py
combinat/root_system/weyl_group.py
combinat/sf/sf.py
groups/matrix_gps/general_linear.py
groups/matrix_gps/special_linear.py
groups/perm_gps/permgroup_named.py
misc/classcall_metaclass.py
misc/constant_function.py
misc/nested_class_test.py
sets/disjoint_union_enumerated_sets.py
sets/finite_enumerated_set.py
sets/non_negative_integers.py
sets/primes.py
structure/all.py
structure/dynamic_class.py
structure/unique_representation.py

[zimmermann6]

comment:6

I tried the patch and sage -t * and got the same failures that I get without the patch
(on x86_64 Fedora12, see #7773). Thus a positive review for me.