GroupCount order by

Question

The Unit test code from the simplegraph-core testsuite below displays the region count of airports but it is not ordered as I would have expected.

The result starts with:

NZ-BOP=  3
MZ-A=  1
MZ-B=  1
IN-TN=  5
MZ-N=  1
PW-004=  1
MZ-I=  2
BS-FP=  1
IN-TR=  1
MZ-T=  1
BJ-AQ=  1
GB-ENG= 27

I looked into

• Gremlin group by vertex property and get sum other properties in the same vertex
• Gremlin query for groupcount of last value

and searched for "GroupCount" in questions tagged gremlin to no avail

What is necessary to fix the ordering?

Unit Test see also https://github.com/BITPlan/com.bitplan.simplegraph/blob/master/simplegraph-core/src/test/java/com/bitplan/simplegraph/core/TestTinkerPop3.java

  @Test
  public void testSortedGroupCount() throws Exception {
    Graph graph = getAirRoutes();
    GraphTraversalSource g = graph.traversal();
    Map<Object, Long> counts = g.V().hasLabel("airport").groupCount()
        .by("region").order().by(Order.decr).next();
    assertEquals(1473, counts.size());
    for (Object key : counts.keySet()) {
      System.out.println(String.format("%s=%3d", key, counts.get(key)));
    }
  }

Morlee

Answer 1

You need to order the values with local scoping:

g.V().hasLabel("airport").
  groupCount().
    by("region").
  order(local).
    by(values, Order.decr)

With local scoping you order within the current traverser (i.e. order the contents of each Map in the traversal).

  @Test
  public void testSortedGroupCount() throws Exception {
    Graph graph = getAirRoutes();
    GraphTraversalSource g = graph.traversal();
    Map<Object, Long> counts = g.V().hasLabel("airport").groupCount()
        .by("region").order(Scope.local).by(Column.values,Order.decr).next();
    // https://mcmap.net/q/2002832/-groupcount-order-by
    assertEquals(1473, counts.size());
    assertEquals("LinkedHashMap",counts.getClass().getSimpleName());
    debug=true;
    if (debug)
      for (Object key : counts.keySet()) {
        System.out.println(String.format("%s=%3d", key, counts.get(key)));
      }

  }

will then show:

US-AK=149
AU-QLD= 50
CA-ON= 46
CA-QC= 44
PF-U-A= 30
US-CA= 29

Matronage

This is the groovy solution. I tried to translate this to java and only found the Scope.local for the order(...) part. for by() I didn't find a two parameter function that fits your answer. Only doing have of your proposal leads to a class cast exception. See the modified testcase I just checked in. – Morlee

it's not a "Groovy" answer really. Static imports are a construct of Java as well as groovy so there is nothing particular to groovy there. you need org.apache.tinkerpop.gremlin.structure.Column for values. Here's the overload of by() you would be using: tinkerpop.apache.org/javadocs/current/core/org/apache/tinkerpop/… – Matronage

I have almost this exact example in the tutorial/book in case it helps kelvinlawrence.net/book/Gremlin-Graph-Guide.html#countrydist – Forepleasure

Answer 2

Using the Gremlin language only, you can do it using the unfold method. This may be useful if you need to run your query by the REST API.

g.V()
  .hasLabel("airport")
  .values("region")
  .groupCount()
  .unfold()
  .order()
  .by(values)

Day