Developing Alternate Graph Types

This section is designed to guide developers who wish to write their own graph structures.

All Graphs functions rely on a standard API to function. As long as your graph structure is a subtype of AbstractGraph and implements the following API functions with the given return values, all functions within the Graphs package should just work:

• edges
• Base.eltype
• edgetype (example: edgetype(g::CustomGraph) = Graphs.SimpleEdge{eltype(g)}))
• has_edge
• has_vertex
• inneighbors
• ne
• nv
• outneighbors
• vertices
• is_directed: Note that since Graphs uses traits to determine directedness, is_directed for a CustomGraph type should be implemented with both of the following signatures:
  • is_directed(::Type{CustomGraph})::Bool (example: is_directed(::Type{<:CustomGraph}) = false)
  • is_directed(g::CustomGraph)::Bool
• zero

If the graph structure is designed to represent weights on edges, the weights function should also be defined. Note that the output does not necessarily have to be a dense matrix, but it must be a subtype of AbstractMatrix{<:Real} and indexable via [u, v].

Note on inheriting from AbstractSimpleGraph

Every subtype of AbstractSimpleGraph must return neighbors in ascending order.