SimpleGraph() SimpleDiGraph() should return a new instance

[filchristou]

according to conversation on #40 #40 (comment) users would commonly expect SimpleGraph(mg::MetaGraph) and SimpleDiGraph(mg::MetaGraph) to return a new instance (deep copy) which they can freely modify without repercussions on mg.

[gdalle]

And we should probably add a get_graph function returning the underlying graph object (which may not be a Simple(Di)Graph anymore thanks to #40)

[gdalle]

As of right now these methods do not exist anymore

[gdalle]

Building a Simple(Di)Graph constructor for general graphs should be part of Graphs.jl, not MetaGraphsNext.jl specifically. Closing here

[filchristou]

well, SimpleGraph(mg::MetaGraph) cannot be implemented in Graphs.jl . So if we want the user to be able to use this constructor it should be defined here.

[gdalle]

It doesn't make much sense to me to have a constructor SimpleGraph(mg::MetaGraph), because in the end it would only use interface functions like vertices and edges. So what I had in mind was a SimpleGraph(g::AbstractGraph) constructor, which does belong in Graphs.jl. What do you think?

[filchristou]

I can live with that yes. Users will then have to do something like SimpleGraph(mg.graph). Do you think the users should access fields with dot-notation or should a getgraph function be defined ?

[gdalle]

I can live with that yes. Users will then have to do something like SimpleGraph(mg.graph). Do you think the users should access fields with dot-notation or should a getgraph function be defined ?

Actually no, that's the whole point: since a MetaGraph is a wrapper around any graph object that extends the usual methods like vertices and edges, if SimpleGraph(::AbstractGraph) is defined in the main package, all users have to do is call SimpleGraph(mg) directly

[gdalle]

This will be taken care of by JuliaGraphs/Graphs.jl#98