FIX take into account mask in the ordering in grid_to_graph

[bthirion]

Reference Issues/PRs

Fixes #18963

What does this implement/fix? Explain your changes.

In grid_to_graph , index the vertices according to mask structure, not their occurrence in edge matrix.

Any other comments?

[glemaitre]

Apart from an additional test, LGTM

Please add an entry to the change log at doc/whats_new/v*.rst. Like the other entries there, please reference this pull request with :pr: and credit yourself (and other contributors if applicable) with :user:.

[ateffal]

Effectively, edges.ravel() (line 80 ) does not contain isolated vertices (vertex of an edge, that was not eliminated by the mask ) while np.where(mask.ravel()) include all vertices.

This is the consequence of the logical and in line 71 : if one of the 2 vertices of an edge is eliminated by the mask, then both are eliminated in the edges array (after line 73).

[bthirion]

Please add an entry to the change log at doc/whats_new/v*.rst. Like the other entries there, please reference this pull request with :pr: and credit yourself (and other contributors if applicable) with :user:.

Shall I create a v1.0.rst ? that would be amazing !

[glemaitre]

Shall I create a v1.0.rst ? that would be amazing !

You don't need to create the file, only to merge master into your branch (you will just have to remove v1.00.rst.
And yep, 1.0 will be out in 6 months hopefully :)

[glemaitre]

LGTM (only a nitpick regarding the what's new entry).

[dPys]

Another related improvement could be a thresholding parameter in the case of an img that zeros out connections for edge weights below a certain value?

[bthirion]

Sorry, I don't understand the Changelog failure... Can anybody help ?

[cmarmo]

@bthirion, do you mind synchronizing with upstream? My typo in the workflow has been fixed in main. This should fix the failure. Sorry for that and thanks for your collaboration.

[bthirion]

Working, thx !

[bthirion]

Any second approval ? ;-)

[TomDLT]

Suggested change

	order = np.searchsorted(np.unique(np.where(mask.ravel())),
	np.arange(maxval + 1))
	order = np.searchsorted(np.flatnonzero(mask), np.arange(maxval + 1))

More readable and probably faster.

[adrinjalali]

@bthirion mind accepting the last comment and merging with main?

[bthirion]

Should be OK now. Best,

[glemaitre]

@adrinjalali do you want to have a look at this one now that the merge conflict is solved.

[glemaitre]

Actually, @TomDLT already reviewed as well. Merging then. Thanks @bthirion