prov.graph

prov.graph converts between a ProvDocument and a NetworkX MultiDiGraph, one node per element (entity, activity, agent) and one edge per relation, so graph algorithms NetworkX provides (and prov does not implement) can run directly over a provenance graph. See Convert to/from a NetworkX graph for worked examples.

prov.graph.prov_to_graph(prov_document: ProvDocument) → nx.MultiDiGraph[Any]

Convert a ProvDocument to a MultiDiGraph instance of the NetworkX library.

The document is first unified() so that records referring to the same real-world thing are merged before the graph is built. Every PROV element becomes a graph node. Each relation becomes an edge between the elements named in its first two formal attributes, with the relation record stored under the edge’s "relation" attribute; if either of those two elements was not already added as a node, a bare node is created for it using INFERRED_ELEMENT_CLASS. Relations whose first two formal attributes are not both populated, or whose attribute type is not in INFERRED_ELEMENT_CLASS, are silently skipped.

Parameters:

prov_document – The ProvDocument instance to convert.

Returns:

A NetworkX MultiDiGraph with PROV elements as nodes and PROV relations as edges.

prov.graph.graph_to_prov(g: nx.MultiDiGraph[Any]) → ProvDocument

Convert a MultiDiGraph that was previously produced by prov_to_graph() back to a ProvDocument.

Every node that is a ProvRecord already attached to a bundle is added to the new document; nodes without a bundle (e.g. bare nodes inferred by prov_to_graph()) are skipped. Every edge whose "relation" data is a ProvRecord is likewise added; edges without relation data are skipped.

Parameters:

g – The graph instance to convert.

Returns:

A new ProvDocument containing the elements and relations recovered from g.