@node A Graph data structure
@section A Graph data structure

This is a simple immutable directed graph.

Most operations are O(n), since Scheme lacks a total order betwen
arbitrary objects.

@defun make-graph [node-key-proc=identity] [node-equal?=eq?]
Constructs a new graph.

@var{node-key-proc} should be a procedure mapping the actual nodes to
a value which can be compared. These values are sometimes called
``node keys''.
@cindex node keys

@var{node-equal?} should return if two nodes are equal or not.
@end defun

@defun rebuild-graph [old-graph] [nodes='()] [edges='()]
@c TODO document me
@end defun

@defun graph-empty? graph
Does the graph contaitn
@end defun

@defun add-node graph node edge-neighbors
Adds the value @var{node} to @var{graph}. @var{edge-neighbors} should
be a list of node keys.
@end defun

@defun get-node graph key
Retrieve a node from the graph, given its node key.

Returns @code{#f} if no such node exists.
@end defun

@defun remove-node graph node
Remvoe @var{node} from @var{graph}, if it exists.
@end defun

@defun find-dangling-node graph
Find a node in the graph which no other node ``depends'' on (has an
edge pointing at it).

NOTE this is O(n^2) (maybe, sort of?)
Getting it faster would require building an index, which
is hard since there isn't a total order on symbols.
@end defun


@defun pop-dandling-node graph
Find a node as per @code{find-dangling-node}, and return two values:
the node, and a new graph without that node.
@end defun

@defun resolve-dependency-graph graph
If each edge is assumed to indicate a nodes dependencies, then this
procedure will find return a flat list where each dependency is before
its dependants.

If any link is missing, or a cycle exists, then the result is
undefined.
@end defun