write a construction for permutation groups so that the coercion system can find a common parent

[mwhansen]

This happens in the coercion branch:

sage: P1 = PermutationGroup([[(1,2)]]); P1
Permutation Group with generators [(1,2)]
sage: P2 = PermutationGroup([[(1,3)]]); P2
Permutation Group with generators [(1,3)]
sage: p1 = P1.gen(); p2 = P2.gen()
sage: p1*p2
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)

/opt/sage-3.0.3.rc0/devel/sage-main/sage/combinat/matrices/<ipython console> in <module>()

/opt/sage-3.0.3.rc0/devel/sage-main/sage/combinat/matrices/element.pyx in sage.structure.element.MonoidElement.__mul__ (sage/structure/element.c:7517)()

TypeError: unsupported operand parent(s) for '*': 'Permutation Group with generators [(1,2)]' and 'Permutation Group with generators [(1,3)]'

Component: group theory

Author: Mike Hansen

Reviewer: David Loeffler

Merged: sage-4.4.4.alpha0

Issue created by migration from https://trac.sagemath.org/ticket/3461

[mwhansen]

Attachment: trac_3461.patch.gz

[mwhansen]

Author: Mike Hansen

[loefflerd]

Attachment: trac_3461_new.patch.gz

replaces previous patch

[loefflerd]

Reviewer: David Loeffler

[loefflerd]

comment:2

The original patch suffered from a particularly evil piece of bitrotting: since the patch was written, some change somewhere has meant that the trivial permutation now has truth value True, not False, which meant that the pushout code would never terminate -- it would keep on returning bigger and bigger towers of (empty) construction functors!

The new patch I've just uploaded gets around this by using the is_one method, instead of a plain truth value test. It passes doctests under 4.4.1. Anyone willing to sign off on my tiny changes, and maybe we can get this into Sage 5.0?

[mwhansen]

comment:3

Looks good to me. Thanks David!

[mwhansen]

Merged: sage-4.4.4.alpha0