The hash function of combinatorial free module elements should depend on the parent

[nthiery]

It would be desirable to make the hash value of (combinatorial) free
module elements depend on the hash value of the parent. However this
currently breaks some code (see
e.g. sage.combinat.crystals.kyoto_path_model.KyotoPathModel).

This code indeed assumes, in dictionary look-ups, that an element of
some free module over ZZ and its counter part over QQ (in the case at
hand, the weight lattice and the weight space), comparing as equal
because of a coercion, have the same hash value.

Yet another case of equality using coercion being harmful:
http://wiki.sagemath.org/EqualityCoercion

Depends on #15931
Depends on #16193

CC: @tscrim @anneschilling

Component: linear algebra

Author: Travis Scrimshaw

Issue created by migration from https://trac.sagemath.org/ticket/15959

[tscrim]

comment:1

Nicolas, do you have a preliminary branch that you could post (or is this the #15931 branch?) or could post the errors/doctest failures?

[nthiery]

comment:3

Replying to @tscrim:

Nicolas, do you have a preliminary branch that you could post (or is this the #15931 branch?) or could post the errors/doctest failures?

Sorry, I should have provided steps to reproduce. Checkout branch #15931 at commit a2bbe7f, and run the tests on the Kyoto path model.

[tscrim]

Author: Travis Scrimshaw

[tscrim]

Commit: 7bdb017

[tscrim]

Branch: u/tscrim/kyoto_fix-15959

[tscrim]

Dependencies: #15931

[tscrim]

comment:4

I haven't merged in the latest #15931, but instead started with the a2bbe7f commit. The fix for the Kyoto path model is that I was passing in elements of the weight space, whereas the Epsilon()/Phi() methods returned an element of the weight lattice. I've added an additional check for robustness.

---

New commits:

e8fe5eb	Trac 15931: implement a proper hash function for (combinatorial) free module elements
e6ef605	Trac 15931: add a @cached_method on the new hash function for (combinatorial) free module elements
a2bbe7f	Trac 15931: include the hash of the parent in the new hash function for (combinatorial) free module elements
13be620	Merge branch 'develop' into test/15959
7bdb017	Fix for Kyoto path model with hash depending on the parent.

[sagetrac-git]

Changed commit from 7bdb017 to bbf4917

[sagetrac-git]

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

2c8ba97	Merge branch 'develop' into u/tscrim/kyoto_fix-15959
bbf4917	Merge branch 'develop' into u/tscrim/kyoto_fix-15959

[sagetrac-git]

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

e778baf

Merge branch 'develop' into u/tscrim/kyoto_fix-15959

[sagetrac-git]

Changed commit from bbf4917 to e778baf

[tscrim]

comment:7

I've merged the branch into #16193 and set that as a dependency.

[tscrim]

Changed dependencies from #15931 to #15931 #16193

[tscrim]

Changed commit from e778baf to none

[tscrim]

Changed branch from u/tscrim/kyoto_fix-15959 to none