an iterator over the leaves of the graph
The leaves are defined as a set of nodes such that every node in the graph is an ancestor of one of the leaves, an no leaf is an ancestor of any other leaf. It is guaranteed to include every node with no children, but one node in a cycle may be chosen if there is no unique leaf for that cycle.
the current implementation will return a minimal leaf set as long as target cycles contain a node with a single child.
add a new link from source to target
If undefined extends LinkDatum, datum can be elided and simply called as
grf.link(source, target);
Rest ...datum: undefined extends LinkDatuman iterator over every link in the graph
add a new node with datum
If undefined extends NodeDatum, datum can be elided and simply called as
grf.node();
Rest ...datum: undefined extends NodeDatuman iterator over every node in the graph
Be careful not to modify the graph structure while iterating as any
modification, including adding or removing links, can change the behavior
of iteration giving unexpected results. If you want to guarantee
consistent iteration, allocating an array first with [...graph.nodes()] will
ensure consistent iteration.
an iterator over the roots of the graph
The roots are defined as a set of nodes such that every node in the graph is a descendant of one of the roots, an no root is a descendant of any other root. It is guaranteed to include every node with no parents, but one node in a cycle may be chosen if there is no unique root for that cycle.
the current implementation will return a minimal root set as long as source cycles contain a node with a single parent.
split a graph into connected components
splits an iterable over a single node in each connected component
Since each node behaves like a graph of its connected component, this is equivalent with returning a graph of each connected component.
compute a topological order of the graph
If the graph can't be represented in topological order, this will try to minimize the number of edge inversions. Optimally minimizing inversions is np-complete, so this will only be approximate.
You can optionally specify a Rank accessor that defines a numerical rank for every node. Nodes with a lower rank will come before nodes of a higher rank even if that requires more edge inversions. Nodes without a rank are unconstrained.
Optional rank: Rank<NodeDatum, LinkDatum>Generated using TypeDoc
a mutable graph representation
In addition to all of the methods inherent to Graphs, this also allows structure graph modification with the node and link methods to either create an empty node or create a link between two nodes. Nodes and links can be removed with the delete method on nodes and links.
Performance
In order to keep track of caches, internally adding links runs union-find which has near, but not quite, constant complexity. However, various methods can run in linear time making some access patterns run in worst case quadratic if they're called in between modifications. This was ultimately deemed acceptable as all modification will likely come before a layout, and that will always take linear time.