Using the Graph Optimization Options

The optimization strategy determines whether the graph is optimized for read-intensive scenarios or for faster updates. It impacts the performance characteristics of graph operations such as querying and updating.

The supported graph optimization options are:

• ReadGraphOption.optimizeFor(GraphOptimizedFor.READ): You can choose this option when the primary operations on the graph are read-based, and updates are infrequent or non-existent.

  When this strategy is selected, the graph's data structures are replicated whenever a new graph or graph snapshot is created. This results in faster and more efficient read operations, such as traversals and queries. However, creating snapshots can be expensive and it can also lead to performance overhead and increased memory usage during updates.

• ReadGraphOption.optimizeFor(GraphOptimizedFor.UPDATES): You can choose this option when the graph is expected to undergo frequent updates, such as adding or removing nodes and edges, or modifying properties of existing elements.

  When this strategy is selected, delta logs are used to manage updates. This approach makes updates more memory-efficient and faster. However, there may be some time overhead when querying the graph due to the need to apply these delta updates.

• ReadGraphOption.synchronizable(): You can choose this option when the graph is expected to be synchronized.

It is important to note the following:

• synchronizable() option can be used in combination with UPDATE and READ. However, the UPDATE and READ options cannot be used at the same time.
• If you are loading a PGQL property graph for SYNCHRONIZABLE option, then ensure that the vertex and edge keys are numeric and non-composite.

The following example loads a PGQL property graph for READ and SYNCHRONIZABLE options:

opg4j> var graph = session.readGraphByName("BANK_GRAPH", GraphSource.PG_PGQL,
...>                            ReadGraphOption.optimizeFor(GraphOptimizedFor.READ),
...>                            ReadGraphOption.synchronizable())
graph ==> PgxGraph[name=BANK_GRAPH_2,N=1000,E=5001,created=1648457198462]

PgxGraph graph = session.readGraphByName("BANK_GRAPH", GraphSource.PG_PGQL,
                                                  ReadGraphOption.optimizeFor(GraphOptimizedFor.READ),
                                                  ReadGraphOption.synchronizable());

>>> graph = session.read_graph_by_name('BANK_GRAPH',
...      'pg_pgql', options=['optimized_for_read', 'synchronizable'])

Using the OnMissingVertex Options

If either the source or destination vertex or both are missing for an edge, then you can use the OnMissingVertexOption which specifies the behavior for handling the edge with the missing vertex. The following values are supported for this option:

• ReadGraphOption.onMissingVertex(OnMissingVertex.ERROR): This is the default option and this specifies that an error must be thrown for edges with missing vertices.
• ReadGraphOption.onMissingVertex(OnMissingVertex.IGNORE_EDGE): Specifies that the edge for a missing vertex must be ignored.
• ReadGraphOption.onMissingVertex(OnMissingVertex.IGNORE_EDGE_LOG): Specifies that the edge for a missing vertex must be ignored and all ignored edges must be logged.
• ReadGraphOption.onMissingVertex(OnMissingVertex.IGNORE_EDGE_LOG_ONCE): Specifies that the edge for a missing vertex must be ignored and only the first ignored edge must be logged.

The following example loads the PGQL property graph by ignoring the edges with missing vertices and logging only the first ignored edge. Note, to view the logs, you must update the default Logback configuration file in /etc/oracle/graph/logback.xml and the graph server (PGX) logger configuration file in /etc/oracle/graph/logback-server.xml to log the DEBUG logs. You can then view the ignored edges in /var/opt/log/pgx-server.log file.

opg4j> session.readGraphByName("REGIONS", GraphSource.PG_PGQL,
...>                             ReadGraphOption.onMissingVertex(OnMissingVertex.IGNORE_EDGE_LOG_ONCE))
$7 ==> PgxGraph[name=REGIONVIEW_3,N=27,E=18,created=1655903219910]

PgxGraph graph = session.readGraphByName("REGIONS", GraphSource.PG_PGQL, ReadGraphOption.onMissingVertex(OnMissingVertex.IGNORE_EDGE_LOG_ONCE));

>>> graph = session.read_graph_by_name('REGIONS',
...      'pg_pgql', options=['on_missing_vertex_ignore_edge_log_once'])

This feature allows you to load a PGQL property graph from another user schema into the graph server (PGX). However, ensure that you have READ permission on all the underlying metadata and data tables when loading a PGQL property graph from another schema.

The following example loads a PGQL property graph from the GRAPHUSER schema:

opg4j> var graph = session.readGraphByName("GRAPHUSER", "FRIENDS", GraphSource.PG_PGQL)
graph ==> PgxGraph[name=FRIENDS,N=6,E=4,created=1672743474212]

PgxGraph graph = session.readGraphByName("GRAPHUSER", "FRIENDS", GraphSource.PG_PGQL);