Use rewrite rules instead of some specialization

[nobrakal]

Hi,

Trying to improve the MaybeNonEmpty version, I discovered that the rewrite rules are very powerful for edgeList (it comes from the quickness of AdjacencyIntMap):

edgeList

Description: Produce a list of the edges in the graph

Mesh

	100
Alga	77.91 μs
AlgaOld	206.3 μs

Clique

	100
Alga	2.347 ms
AlgaOld	7.619 ms

Circuit

	100
Alga	39.63 μs
AlgaOld	118.9 μs

SUMMARY:

• Alga was the fastest 3 times

ABSTRACT:
(Based on an average of the ratio between largest benchmarks)

• Alga was 2.94 times faster than AlgaOld

[snowleopard]

edgeCount: 0.41 (good)
edgeList: 0.32 (good)

@nobrakal Awesome, thank you!

Will the same come for non-empty graphs?

[nobrakal]

Will the same come for non-empty graphs?

I think yes, as well as many SPECIALIZE and other RULES of the Algebra.Graph module :)

So since we make no use for now of rules pragma in Algebra.Graph.NonEmpty, I don't know if it worth it adding only one or two here.

(In fact, I already added almost all of these rules in Algebra.Graph.NonEmpty for my work related to #90 )

[snowleopard]

Right, I see. We could split the work on non-empty graphs into a separate PR.

as well as many SPECIALIZE and other RULES of the Algebra.Graph module :)

Are these essentially the same as in this PR? In this case, I think it makes sense to combine all such changes together in one PR. Do you agree? If yes, I'll wait for further commits.

[nobrakal]

Are these essentially the same as in this PR?

Yep this is mainly specialize and rules pragma. In fact, it would be great to add these pragmas directly into the ToGraph class when possible, so it benefit other implementations. What do you think ?

[snowleopard]

If you manage to do this at the ToGraph class level it would be great, of course. But will that actually work?

[nobrakal]

But will that actually work?

You were right, I was a bit optimistic... I as not able to sneak this into the ToGraph class (I think this is due to TypeFamilies, which make the simple type-match of rewrite rules a bit complicated).

So I think it is a good idea to open an other PR concerning non-empty graphs, since there will be some different work unrelated to edgeList

[snowleopard]

OK, sounds good. Then I'm waiting for further Graph-related commits.

[nobrakal]

I'm waiting for further Graph-related commits.

I do not have any :) Do you see anything missing for Algebra.Graph ?

[snowleopard]

Ah, I think I just misunderstood one of your earlier messages :) Merged now!