Tutorial Instructions

Here you will find the tasks of the Gradoop tutorial.

Lessons
Subgraph
Grouping
Pattern Matching
Complex social network analysis

---

Lesson 1 - The subgraph operator

Task 1 - Filter by label

• Create a subgraph that contains ‘person’ vertices and ‘knows’ edges only.

Solution

Use the ByLabel class to filter for vertices/edges of a specific label.

graph = graph.subgraph(new ByLabel<>("person"), new ByLabel<>("knows"));

Task 2 - Logical operators

• Get universities and persons that used the browser Firefox.
• Include all relationships between these vertices.

Some help

• Use the function .vertexInducedSubgraph() to only filter for vertices and get all edges between them.
• Use the ByLabel or LabelIsIn classes to filter for vertices/edges of a specific label.
• Use the ByProperty class to filter for elements with a specific property and optional a value.
• Use the And and Or classes to build boolean relations.

Solution

graph = graph.vertexInducedSubgraph(
    new Or<>(
      new ByLabel<>("university"),
      new And<>(new ByLabel<>("person"), new ByProperty<>("browserUsed", PropertyValue.create("Firefox")))));

Task 3 - User-defined filter

• Some edge types have a creationDate property. Get all edges that contain this property and are crated in the period [2012-01-01 00:00 , 2012-01-05 00:00) by using a user-defined edge filter function.
• Only the source and target vertices of these edges should be included.
• Optional: Use parameters (by adding a constructor) to define the lower and upper bound for the condition.

Some help

• Use the function .edgeInducedSubgraph() to only filter for edges and get all connected vertices.
• You can check the availability of a property by edge.hasProperty("myProperty").
• You get the value of a propety by calling edge.getPropertyValue("myProperty").get{dataType}()

Solution

This is the solution of the task with the optional subtask included.

graph = graph.edgeInducedSubgraph(
    new CreatedInPeriod<>(LocalDateTime.of(2012, 1, 1, 0, 0), LocalDateTime.of(2012, 1, 5, 0, 0)));

private static class CreatedInPeriod<E extends Edge> implements FilterFunction<E> {

  private LocalDateTime lowerBound, upperBound;

  public CreatedInPeriod(LocalDateTime start, LocalDateTime end) {
    this.lowerBound = start;
    this.upperBound = end;
  }

  @Override
  public boolean filter(E edge) throws Exception {
    if (edge.hasProperty("creationDate")) {
      LocalDateTime creationDate = edge.getPropertyValue("creationDate").getDateTime();
      return (creationDate.isEqual(lowerBound) || creationDate.isAfter(lowerBound)) &&
        creationDate.isBefore(upperBound);
    }
    return false;
  }
}

Lesson 2 - The Grouping operator

Have a look at the Wiki pages of Gradoop for further details about this operator. There are several pre-defined aggregations available.

Task 1 - Schema graph

• Create a schema graph by grouping vertices and edges by the label.
• How many vertices and edges exist per label?

Some help

• The class GroupingKeys offers pre-defined key functions.

Solution

graph = graph.callForGraph(
    new KeyedGrouping<>(
      Arrays.asList(GroupingKeys.label()),
      Arrays.asList(new Count()),
      Arrays.asList(GroupingKeys.label()),
      Arrays.asList(new Count())
    )
  );

Task 2 - Attributed grouping

• How many males and females are there? What is the average age per gender?
• How many males know women and vice versa?
• How many male and female people study at universities per class year?

You already have a subgraph with person and university nodes and knows and studyAt edges.

All person vertices have now an additional property age as Integer.

Some help

• Use GroupingKeys.property("myProperty") to group on a property.
• Use the aggregate new AverageProperty("myProperty") to calculate the average of a numerical value.

Solution

graph = graph.callForGraph(
    new KeyedGrouping<>(
      Arrays.asList(GroupingKeys.label(), GroupingKeys.property("gender")),
      Arrays.asList(new Count(), new AverageProperty("age")),
      Arrays.asList(GroupingKeys.label(), GroupingKeys.property("classYear")),
      Arrays.asList(new Count())
    )
  );

Task 3 - User-defined key function

• How many people are born on the same weekday (Mon – Sun)? You have to create a UDF that extracts the day from the property birthday.
• How old is the youngest and oldest person in the group?
• How many know each other from these groups?

You already have a subgraph with person nodes and knows edges.

All person vertices have now an additional property age as Integer.

Some help

• Use the aggregate new MinProperty("myProperty") to calculate the minimum of a numerical value.
• Use the aggregate new MaxProperty("myProperty") to calculate the maximum of a numerical value.
• You can check the availability of a property by vertex.hasProperty("myProperty").
• You get the value of a propety by calling vertex.getPropertyValue("myProperty").get{dataType}()
• The Class LocalDate offers a function .getDayOfWeek().name() to get the name of the weekday as string.

Solution

graph = graph.callForGraph(
    new KeyedGrouping<>(
      Arrays.asList(GroupingKeys.label(), new GetYearOfDate<>()),
      Arrays.asList(new Count(), new MinProperty("age"), new MaxProperty("age")),
      Arrays.asList(GroupingKeys.label()),
      Arrays.asList(new Count())
    )
  );

private static class GetYearOfDate<V extends Vertex> implements KeyFunction<V, String> {

  @Override
  public String getKey(V vertex) {
    String dayOfWeek = "unknown";
    if (vertex.hasProperty("birthday")) {
      dayOfWeek = vertex.getPropertyValue("birthday").getDate().getDayOfWeek().name();
    }
    return dayOfWeek;
  }

 ...
}

Pattern Matching

Complex social network analysis