Matrix and Components should have a sparse iterator

[mwageringel]

As observed on this Ask SageMath question, it does not seem to be possible to iterate over the non-zero entries of a tensor. This is unfortunate because the entries are stored in a sparse format, in a dictionary.

Since a tensor might have symmetries, this is more involved than just iterating over the dictionary, but such an iterator would immediately be useful for the implementation of the display() method, for instance.

Here we implement a method Components.items, returning an iterable of (indices, value). This is compatible with the sparse and dense vectors from sage.modules.

(Because a Components does not currently have a parent, it does not make sense to define the method monomial_coefficients - as defined by ModulesWithBasis.)

We also define Matrix.items with the same interface.

CC: @egourgoulhon @LBrunswic @honglizhaobob @tscrim

Component: linear algebra

Author: Matthias Koeppe

Branch/Commit: 74d9493

Reviewer: Eric Gourgoulhon

Issue created by migration from https://trac.sagemath.org/ticket/29619

[egourgoulhon]

comment:2

Thanks for opening this ticket. I've added a link to it in the "Algebraic part" section of the meta-ticket #18528.

[mkoeppe]

comment:3

See also: #30309 - Unify free module elements API: methods dict, monomial_coefficients, etc.

[mkoeppe]

comment:5

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

[mkoeppe]

comment:7

@honglizhaobob This is probably the best ticket for starting

[honglizhaobob]

comment:8

Replying to @mkoeppe:

@honglizhaobob This is probably the best ticket for starting

Is there a set of unit test cases one needs to pass to ensure a justified change of data structure?

[mkoeppe]

comment:9

To clarify - here on this ticket, the data structure shouldn't be changed but instead methods added.

Sage mainly uses doctests - see https://doc.sagemath.org/html/en/developer/doctesting.html
but there are also unit tests, which are invoked within the doctests by TestSuite(...).run().

[honglizhaobob]

Branch: u/gh-honglizhaobob/tensors_should_have_a_sparse_iterator

[sagetrac-git]

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

bbed88a

added doc for iter

[sagetrac-git]

Commit: bbed88a

[mkoeppe]

comment:12

+        for multi_idx in self._comp.keys():
+            # access value in dictionary
+            coef = self._comp[multi_idx]

You can use .items() here - see https://docs.python.org/3/tutorial/datastructures.html#looping-techniques

[mkoeppe]

Changed branch from u/gh-honglizhaobob/tensors_should_have_a_sparse_iterator to u/mkoeppe/tensors_should_have_a_sparse_iterator

[mkoeppe]

New commits:

3862383	src/sage/tensor/modules/comp.py: WIP
882c5ea	Components.items: New

[mkoeppe]

Changed commit from bbed88a to 882c5ea

[mkoeppe]

Author: Matthias Koeppe

[sagetrac-git]

Changed commit from 882c5ea to 4958211

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

4958211

Components.items: New

[sagetrac-git]

Changed commit from 4958211 to 74d9493

[sagetrac-git]

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

74d9493

Matrix.items: New

[egourgoulhon]

comment:23

Thanks for implementing this. LGTM.

[egourgoulhon]

Reviewer: Eric Gourgoulhon

[mkoeppe]

comment:24

Thanks!

[vbraun]

Changed branch from u/mkoeppe/tensors_should_have_a_sparse_iterator to 74d9493