Fix summary row issues in case postaggregations are happening

[kgyrtkirk]

• fix issue with summary rows (Fix summary row issues in case postaggregations are happening #15232):
  after the summary row fixes of Count distinct returned incorrect results without useApproximateCountDistinct #14748 ; queries which were computing a postaggregate-s in an empty summary row have encountered AIOOBE-s; because the post aggregation columns were not allocated correctly.

In case the table has no negative values for m1; the following query could produce such an exception:

select count(distinct m1)*2 from druid.foo where m1 < -1.0

• add suggerstion to enable windowing if the query uses that feature

[clintropolis]

oh, I think you need to actually compute the postaggs... which is why the test is failing in default value mode, see mergeResults for an example, something like

  private static Iterator<ResultRow> summaryRowIterator(GroupByQuery q)
  {
    List<AggregatorFactory> aggSpec = q.getAggregatorSpecs();
    ResultRow resultRow = ResultRow.create(q.getResultRowSizeWithPostAggregators());
    for (int i = 0; i < aggSpec.size(); i++) {
      resultRow.set(i, aggSpec.get(i).factorize(new AllNullColumnSelectorFactory()).get());
    }
    Map<String, Object> map = resultRow.toMap(q);
    for (int i = 0; i < q.getPostAggregatorSpecs().size(); i++) {
      final PostAggregator postAggregator = q.getPostAggregatorSpecs().get(i);
      final Object value = postAggregator.compute(map);

      resultRow.set(q.getResultRowPostAggregatorStart() + i, value);
      map.put(postAggregator.getName(), value);
    }
    return Collections.singleton(resultRow).iterator();
  }

makes the test pass in both modes for me

[kgyrtkirk]

thank you @clintropolis for taking a look and for the suggested fix :)