py3: fix doctests in geometry/cone

[fchapoton]

by sorting

CC: @orlitzky @novoselt

Component: python3

Author: Frédéric Chapoton

Branch/Commit: 65f4f7b

Reviewer: Andrey Novoseltsev

Issue created by migration from https://trac.sagemath.org/ticket/26636

[fchapoton]

New commits:

65f4f7b

py3: fix doctests in geometry/cone

[fchapoton]

Branch: u/chapoton/26636

[fchapoton]

Commit: 65f4f7b

[orlitzky]

comment:3

For what it's worth, I believe the non-determinism is due to the following in semigroup_generators:

    gens = set()
    for cone in subcones:
        gens.update(cone.semigroup_generators())
    return tuple(gens)

An ordered tuple or PointCollection (which amounts to the same thing) is always returned, but using set() to eliminate the duplicates makes the order of the tuple itself unreliable.

Andrey will know whether this should be fixed by sorting the output, or by making the output deterministic.

[novoselt]

comment:4

Only sorting the output - there is no canonical ordering, so there is no point in wasting time on sorting it always.

[novoselt]

Reviewer: Andrey Novoseltsev

[vbraun]

Changed branch from u/chapoton/26636 to 65f4f7b