sage.tensor.modules.free_module_basis: Make Basis_abstract a subclass of AbstractFamily

[mkoeppe]

... in particular provide an implementation of the keys() method.

We also add a generic method AbstractFamily.items() to complement the methods keys() and values(), mimicking the Mapping protocol. (Split out from #34340.)

CC: @mjungmath @egourgoulhon @tscrim

Component: linear algebra

Author: Matthias Koeppe

Branch/Commit: 66874a9

Reviewer: Eric Gourgoulhon

Issue created by migration from https://trac.sagemath.org/ticket/30300

[mkoeppe]

comment:2

Any help with implementing this would be very welcome

[mkoeppe]

comment:3

Setting new milestone based on a cursory review of ticket status, priority, and last modification date.

[mkoeppe]

Changed dependencies from #30287, #30259 to #30287, #30259, #34340

[mkoeppe]

Changed dependencies from #30287, #30259, #34340 to #30287, #34340

[mkoeppe]

Branch: u/mkoeppe/sage_tensor_modules_free_module_basis__make_basis_abstract_a_subclass_of_family

[mkoeppe]

New commits:

78400bc	AbstractFamily.{keys,values,items}: New
d2e4466	src/sage/sets/family.py: Add missing import
d15bd7d	src/sage/tensor/modules/free_module_basis.py: Make Basis_abstract a subclass of AbstractFamily

[mkoeppe]

Commit: d15bd7d

[mkoeppe]

Author: Matthias Koeppe

[mkoeppe]

Changed dependencies from #30287, #34340 to none

[sagetrac-git]

Changed commit from d15bd7d to eeba29a

[sagetrac-git]

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

eeba29a

src/sage/tensor/modules/free_module_basis.py: Add documentation

[mkoeppe]

comment:14

sage -t --random-seed=263111727722887782759446361100915924402 sage/manifolds/differentiable/chart.py  # 2 doctests failed
sage -t --random-seed=263111727722887782759446361100915924402 sage/manifolds/differentiable/vectorframe.py  # 2 doctests failed
sage -t --random-seed=263111727722887782759446361100915924402 sage/manifolds/local_frame.py  # 2 doctests failed

[sagetrac-git]

Changed commit from eeba29a to 5936c6b

[sagetrac-git]

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

5936c6b

src/sage/manifolds: Update doctest outputs

[egourgoulhon]

comment:16

Could you please remind me what is the purpose of this ticket? In particular, why should bases of a FiniteRankFreeModule have a keys() method?

[sagetrac-git]

Changed commit from 5936c6b to 66874a9

[sagetrac-git]

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

66874a9

src/sage/sets/family.py: Add doctest

[egourgoulhon]

comment:21

Replying to @mkoeppe:

Replying to @egourgoulhon:

Could you please remind me what is the purpose of this ticket? In particular, why should bases of a FiniteRankFreeModule have a keys() method?

By itself, this change only improves "discoverability": If a user has an instance of Basis_abstract, currently they can get an element by calling __getitem__ and either get an element or an error. (Or they would have to know the detail that the indices accepted by __getitem__ are the underlying modules's irange.)

It is also preparation for #30229, which makes the standard bases of tensor modules (induced by the bases of their base module) explicit objects (instances of the subclass TensorFreeSubmoduleBasis_comp), instead of just being implicit in various algorithms.

Thanks for your answer. It's still not clear to me what the advantage of making Basis_abstract a subclass of AbstractFamily is. It adds some code complexity, making FreeModuleBasis be a Parent, but I trust you it is worth it. Could you tell which functionality of AbstractFamily or FiniteEnumeratedSets (the declared category) do you plan to use in #30229?

[mkoeppe]

comment:23

Replying to @egourgoulhon:

Could you tell which functionality of AbstractFamily or FiniteEnumeratedSets (the declared category) do you plan to use in #30229?

The new version of method isomorphism_with_fixed_basis in #34427 illustrates what I have in mind

[egourgoulhon]

Reviewer: Eric Gourgoulhon

[egourgoulhon]

comment:24

Replying to @mkoeppe:

Replying to @egourgoulhon:

Could you tell which functionality of AbstractFamily or FiniteEnumeratedSets (the declared category) do you plan to use in #30229?

The new version of method isomorphism_with_fixed_basis in #34427 illustrates what I have in mind

It looks that this uses only the method keys(), which can be defined without any reference to AbstractFamily. However, I understand that the inheritance from AbstractFamily makes the bases of FiniteRankFreeModule's more similar to those of other free modules in Sage, which is certainly a good thing => positive review from my side.

[mkoeppe]

comment:25

Thank you!

[mkoeppe]

comment:26

(commented on the wrong ticket)

[mkoeppe]

Dependencies: #31276

[mkoeppe]

Changed dependencies from #31276 to none

[vbraun]

Changed branch from u/mkoeppe/sage_tensor_modules_free_module_basis__make_basis_abstract_a_subclass_of_family to 66874a9