Using the DseGraphFrame framework for graph analytics queries

The DseGraphFrame framework allows you create applications that use the Spark API for analytics operations on DSE Graph. It is inspired by the Databricks GraphFrame library and supports a subset of the Gremlin graph traversal language. You can read DSE Graph data into a GraphFrame and write GraphFrame objects from any format supported by Spark into DSE Graph. You can also query DseGraphFrame vertices and edges in Spark SQL.

Choosing when to use DseGraphFrame or DSE Graph OLAP queries

DSE Graph OLAP has broader support for Gremlin than the DseGraphFrame API. While Graph OLAP is the best choice for deep queries, simple filtering and counts are much faster using the DseGraphFrame API.

Overview of DseGraphFrame

DseGraphFrame represents a graph as two virtual tables: a vertex and an edge DataFrame. The V() method returns the vertex DataFrame of a graph. The E() method returns the edge DataFrame of a graph.

val g = spark.dseGraph("test")

DseGraphFrame uses a GraphFrame-compatible format. This format requires the vertex DataFrame to have only one id column and the edge DataFrame to have hard coded src and dst columns. Since DSE Graph allows users to define any arbitrary set of columns as the vertex id and since there is no concept of labels in GraphFrame, DseGraphFrame will serialize the entire DSE Graph id into one id column. The label is represented as part of the id and also as the ~label property column.

Using DseGraphFrame

The starting point for all operations is the DseGraphFrame object. In Scala, there’s an implicit conversion between DseGraphFrame objects and GraphFrame objects.

// load a graph
val graph = spark.dseGraph("my_graph")
//use the TinkerPop API
graph.V().has("edge", gt(100)).count().next()
// use the GraphFrame API
graph.find("(a)-[e]->(b); (b)-[e2]->(c)").filter("e2.`~label` = 'includedIn'").select("", "e.`~label`", "", "e2.`~label`", "")
// Use both the TinkerPop and GraphFrame APIs:

In Java, use the gf() method, or use the DseGraphFrameBuilder.dseGraph(String graphName, GraphFrame gf) method to return a GraphFrame instance.

//load a graph
GraphFrame graph = DseGraphFrameBuilder.dseGraph("my_graph", spark);
//use the TinkerPop API
graph.V().has("edge", gt(100)).count().next()
// use the GraphFrame API
graph.find("(a)-[e]->(b); (b)-[e2]->(c)").filter("e2.label = 'includedIn'").select("", "e.`~label`", "", "e2.`~label`", "").distinct().show()
// Use both the TinkerPop and GraphFrame APIs:

Before doing complex queries, it is strongly recommended you cache the graph. You can do so using the cache() or persist(level) methods.


The persist() method requires one of the Spark persist levels as a parameter.


To list all graph names, call the spark.dseGraphs method.


See the API documentation for a full list of methods.

Copying graphs from one cluster to another using DseGraphFrame

You can copy graph data from one DSE cluster to another using DseGraphFrame. Modify the configuration in the Spark session to include both cluster names and connection points.

spark.setCassandraConf("cluster1", CassandraConnectorConf.ConnectionHostParam.option(""))
spark.setCassandraConf("cluster2", CassandraConnectorConf.ConnectionHostParam.option(""))

Connect to the graphs on the source and destination clusters.

val source =  spark.dseGraph("srcGraph", Map ("cluster" -> "cluster1") )
val dst =  spark.dseGraph("dstGraph", Map ("cluster" -> "cluster2"))

Update the vertices and edges on the destination graph using the source graph.


Using authorization with DseGraphFrame

If you have enabled authorization on DSE, grant execute permissions to the DseGraphRpc object.

TinkerPop API support in DseGraphFrame

DseGraphFrame supports a subset of the Apache TinkerPop traversal API.

Mapping rules for DseGraphFrame

DseGraphFrame uses mapping rules for column names and types.

Exporting graphs using DseGraphFrame

Use DseGraphFrame to export the graph to any format supported by Spark.

Importing graphs using DseGraphFrame

Use DseGraphFrame to import a graph to DataStax Enterprise.

Was this helpful?

Give Feedback

How can we improve the documentation?

© 2024 DataStax | Privacy policy | Terms of use

Apache, Apache Cassandra, Cassandra, Apache Tomcat, Tomcat, Apache Lucene, Apache Solr, Apache Hadoop, Hadoop, Apache Pulsar, Pulsar, Apache Spark, Spark, Apache TinkerPop, TinkerPop, Apache Kafka and Kafka are either registered trademarks or trademarks of the Apache Software Foundation or its subsidiaries in Canada, the United States and/or other countries. Kubernetes is the registered trademark of the Linux Foundation.

General Inquiries: +1 (650) 389-6000,