Make behaviour of DigraphRemoveEdge{s} consistent when removing no edges

[Peter-Ing]

Added an additional method to preserve properties in the case of empty removals.

This addresses issue #260. (Note from Wilf: currently only partially.)

[wilfwilson]

Please can you also add a test for this behaviour, which verifies that the digraph return by DigraphRemoveEdges, when given an immutable digraph D and an empty list, is indeed the identical object to D.

Please you can you also fixed the spelling mistakes in the commit message, and indeed reformulate the commit message, so that its first line (if possible) consists of 50 or fewer characters?

[wilfwilson]

Oh wait, what is the relationship between this PR and #338?

[Peter-Ing]

Oh wait, what is the relationship between this PR and #338?

this request should have been an update to #338 but as it hasn't updated it and comments have been left here I have closed #338.

[james-d-mitchell]

@Peter-Ing I agree with @wilfwilson. I think the intended changes would be the following:

if digraph is immutable:

• DigraphRemoveEdges(digraph, [ ]) returns digraph
• DigraphRemoveEdge(digraph, [ ]) gives the current error (the argument is not valid)
• DigraphRemoveEdges(digraph, [[src, ran]]) returns an immutable copy of digraph only when [src, ran] is valid. If [src, ran] is not valid, then the function could either give an error, or just return digraph (i.e. not a copy)
• DigraphRemoveEdge(digraph, [src, ran]) returns an immutable copy of digraph only when [src, ran] is valid. If [src, ran] is not valid, then the function could either give an error, or just return digraph (i.e. not a copy)

I'm not sure which is preferable: giving an error or just returning digraph and doing nothing. The latter would be more consistent with the current behaviour, but the former would probably be more helpful. Any thoughts?

[james-d-mitchell]

Thanks @Peter-Ing, I think this PR should be reworked to reflect the comments I made, feel free to leave it open, and just rework your new changes on top of this.

[wilfwilson]

I tend to prefer the former kind of behaviour, I think it is what most people (or at least I) would expect. The latter kind of behaviour seems like it's done for optimisation, letting you reduce code (not having a separate existence check first, in some cases) and save time (by not twice checking whether an edge exists ). Unless users ask for that kind of behaviour, I don't think we should proactively offer it.

[github-actions]

Stale pull request message