Random remarks

[gdalle]

Hey there @SteffenPL,

Taking a very quick look at your code, here are a few things I thought of. We can always structure them later.

• I would separate the various structs into different files to make it clearer which methods belong to which
• Why do you need a new edge type?
• Assertions can be deactivated, better to throw errors for invalid inputs
• Is edge addition type-stable? I'm not sure what the callbacks are for but they could lead to performance issues (see the infamous https://docs.julialang.org/en/v1/manual/performance-tips/#man-performance-captured)

[SteffenPL]

Thanks for looking at the project! I'll implement 1/ and 3/ directly.

About 4/, it seemed to be type-stable here, but I will check it again.

About 2/, you are actually right. My initial idea of the implementation needed a new edge type, but the more I simplified it, the less it was needed anymore and I did not realize that. I can use the simple edge type from Graphs.jl and define a new init_edge function to call minmax in the case of undirected graphs.

[SteffenPL]

Hi, I implemented 1-3.

Regarding the add_edge operation, that seems to be type stable. (The difference to the example from the Julia docs is that the callbacks are only used internally, but they have a finite lifetime. I think Julia seems smart enough not to capture anything here...) Therefore, I think 4 is also fine.

If it would be instable, then also this test would fail:

BoundedDegreeGraphs.jl/test/runtests.jl

Line 158 in 22443f7

@testset "allocations with edge data" begin