-
Notifications
You must be signed in to change notification settings - Fork 239
User Defined Steps
Gremlin provides the ability for a user to define their own step definitions natively in Groovy or in Java. This is very useful when wishing to work with your low-level graph data at a higher level of abstraction. This section will discuss how to write your own step definitions and demonstrate how they are useful for making your Gremlin code more concise and more self-explanatory.
Gremlin comes with a collection of built-in step definitions (see Gremlin Steps). It is possible for developers to create their own step definitions. Simply add a closure that represents the step to the respective classes.
c = {_{x = it}.outE[[label:'created']].inV.inE[[label:'created']].outV{x != it}}
[Iterable,Iterator,Vertex].each{it.metaClass.codeveloper = { Gremlin.compose(delegate, c()) }}
In the code above, c
is a closure to create the desired composite step and the second line is required to add c
as the co_developer
step. Given the graph diagrammed in Defining a Property Graph, we can determine the co-developers of a particular vertex.
gremlin> g = TinkerGraphFactory.createTinkerGraph()
==>tinkergraph[vertices:6 edges:6]
gremlin> g.v(1).codeveloper
==>v[4]
==>v[6]
Realize that this step definition can be used like any other step definition.
gremlin> g.v(1).codeveloper.name
==>josh
==>peter
What step definitions allow you to do is to work with “higher order” relationships in your graph. Thus, instead of working at the level of
_{x = it}.outE[[label:'created']].inV.inE[[label:'created']].outV{x != it}
you can work at the more semantically natural level of
codeveloper
You can also create an anonymous step (or “lambda step”) using the step
step. For example:
gremlin> g = TinkerGraphFactory.createTinkerGraph()
==>tinkergraph[vertices:6 edges:6]
gremlin> g.v(1).outE.inV.step{ starts.next().map }
==>{name=vadas, age=27}
==>{name=lop, lang=java}
==>{name=josh, age=32}
The property starts
is defined in Pipes and represents the previous pipe in the pipeline. For those familiar with Pipes, the provided step
closure becomes the method AbstractPipe.processNextStart()
.
To ensure speed, it is possible to define steps in Java and then load the paths using import
. Here is the previous co-developer
path written in Java as a pipe (see Pipes).
public class CoDeveloperPipe extends Pipeline<Vertex,Vertex> {
public CoDeveloperPipe() {
List singleObjectList = new ArrayList() {
public boolean add(Object object) {
this.clear();
super.add(object);
return true;
}
};
Pipe pipe0 = new AggregatorPipe(singleObjectList);
Pipe pipe1 = new VertexEdgePipe(Step.OUT_EDGES);
Pipe pipe2 = new LabelFilterPipe("created", Filter.NOT_EQUALS);
Pipe pipe3 = new EdgeVertexPipe(Step.IN_VERTEX);
Pipe pipe4 = new VertexEdgePipe(Step.IN_EDGES);
Pipe pipe5 = new LabelFilterPipe("created", Filter.NOT_EQUALS);
Pipe pipe6 = new EdgeVertexPipe(Step.OUT_VERTEX);
Pipe pipe7 = new ObjectFilterPipe(pipe0.getSideEffect(), Filter.EQUALS);
this.setPipes(pipe0, pipe1, pipe2, pipe3, pipe4, pipe5, pipe6, pipe7);
}
}
Make sure that this class is in your Java classpath. If so, you can now include this newly created step/pipe in Gremlin.
[Iterable,Iterator,Vertex].each{ it.metaClass.codeveloper = {Gremlin.compose(delegate, new CoDeveloperPipe())}}