TinkerpopClientService 2.4.0.4.11.0.0-352

Bundle
org.apache.nifi | nifi-other-graph-services-nar
Description
This service interacts with a tinkerpop-compliant graph service, providing both script submission and bytecode submission capabilities. Script submission is the default, with the script command being sent to the gremlin server as text. This should only be used for simple interactions with a tinkerpop-compliant server such as counts or other operations that do not require the injection of custom classed. Bytecode submission allows much more flexibility. When providing a jar, custom serializers can be used and pre-compiled graph logic can be utilized by groovy scriptsprovided by processors such as the ExecuteGraphQueryRecord.
Tags
graph, gremlin
Input Requirement
Supports Sensitive Dynamic Properties
false
  • Additional Details for TinkerpopClientService 2.4.0.4.11.0.0-352

    GremlinClientService

    Description:

    This client service configures a connection to a Gremlin Server and allows Gremlin queries to be executed against the Gremlin Server. For more information on Gremlin and Gremlin Server, see the Apache Tinkerpop project.

    This client service supports two differnt modes of operation: Script Submission and Bytecode Submission, described below.

    Script Submission

    Script submission is the default way to interact with the gremlin server. This takes the input script and uses Script Submission to interact with the gremlin server. Because the script is shipped to the gremlin server as a string, only simple queries are recommended ( count, path, etc.) as there are no complex serializers available in this operation. This also means that NiFi will not be opinionated about what is returned, whatever the response from the tinkerpop server is, the response will be deserialized assuming common Java types. In the case of a Map return, the values will be returned as a record in the FlowFile response, in all other cases, the return of the query will be coerced into a Map with key “result” and value being the result of your script submission for that specific response.

    Serialization Issues in Script Submission

    A common issue when creating Gremlin scripts for first time users is to accidentally return an unserializable object. Gremlin is a Groovy DSL and so it behaves like compiled Groovy including returning the last statement in the script. This is an example of a Gremlin script that could cause unexpected failures:

    g.V().hasLabel("person").has("name", "John Smith").valueMap()

    The valueMap() step is not directly serializable and will fail. To fix that you have two potential options:

    //Return a Map
g.V().hasLabel("person").has("name", "John Smith").valueMap().next()

    Alternative:

    g.V().hasLabel("person").has("name", "John Smith").valueMap()
true //Return boolean literal

    Bytecode Submission

    Bytecode submission is the more flexible of the two submission method and will be much more performant in a production system. When combined with the Yaml connection settings and a custom jar, very complex graph queries can be run directly within the NiFi JVM, leveraging custom serializers to decrease serialization overhead.

    Instead of submitting a script to the gremlin server, requiring string serialization on both sides of the string result set, the groovy script is compiled within the NiFi JVM. This compiled script has the bindings of g (the GraphTraversalSource) and log (the NiFi logger) injected into the compiled code. Utilizing g, your result set is contained within NiFi and serialization should take care of the overhead of your responses drastically decreasing the likelihood of serialization errors.

    As the result returned cannot be known by NiFi to be a specific type, your groovy script must rerun a Map<String, Object>, otherwise the response will be ignored. Here is an example:

    Object results = g.V().hasLabel("person").has("name", "John Smith").valueMap().collect()
[result: results]

    This will break up your response objects into an array within your result key, allowing further processing within nifi if necessary.

Properties