Search queries

A search query returns dimension values that match the search specification.

{
  "queryType": "search",
  "dataSource": "sample_datasource",
  "granularity": "day",
  "searchDimensions": [
    "dim1",
    "dim2"
  ],
  "query": {
    "type": "insensitive_contains",
    "value": "Ke"
  },
  "sort" : {
    "type": "lexicographic"
  },
  "intervals": [
    "2013-01-01T00:00:00.000/2013-01-03T00:00:00.000"
  ]
}

There are several main parts to a search query:

The format of the result is:

[
  {
    "timestamp": "2013-01-01T00:00:00.000Z",
    "result": [
      {
        "dimension": "dim1",
        "value": "Ke$ha",
        "count": 3
      },
      {
        "dimension": "dim2",
        "value": "Ke$haForPresident",
        "count": 1
      }
    ]
  },
  {
    "timestamp": "2013-01-02T00:00:00.000Z",
    "result": [
      {
        "dimension": "dim1",
        "value": "SomethingThatContainsKe",
        "count": 1
      },
      {
        "dimension": "dim2",
        "value": "SomethingElseThatContainsKe",
        "count": 2
      }
    ]
  }
]

Implementation details

Strategies

Search queries can be executed using two different strategies. The default strategy is determined by the “druid.query.search.searchStrategy” runtime property on the Broker. This can be overridden using “searchStrategy” in the query context. If neither the context field nor the property is set, the “useIndexes” strategy will be used.

• “useIndexes” strategy, the default, first categorizes search dimensions into two groups according to their support for bitmap indexes. And then, it applies index-only and cursor-based execution plans to the group of dimensions supporting bitmaps and others, respectively. The index-only plan uses only indexes for search query processing. For each dimension, it reads the bitmap index for each dimension value, evaluates the search predicate, and finally checks the time interval and filter predicates. For the cursor-based execution plan, please refer to the “cursorOnly” strategy. The index-only plan shows low performance for the search dimensions of large cardinality which means most values of search dimensions are unique.

• “cursorOnly” strategy generates a cursor-based execution plan. This plan creates a cursor which reads a row from a queryableIndexSegment, and then evaluates search predicates. If some filters support bitmap indexes, the cursor can read only the rows which satisfy those filters, thereby saving I/O cost. However, it might be slow with filters of low selectivity.

• “auto” strategy uses a cost-based planner for choosing an optimal search strategy. It estimates the cost of index-only and cursor-based execution plans, and chooses the optimal one. Currently, it is not enabled by default due to the overhead of cost estimation.

Server configuration

The following runtime properties apply:

Query context

The following query context parameters apply: