mapper-murmur3.asciidoc

Mapper murmur3 plugin

The mapper-murmur3 plugin provides the ability to compute hash of field values at index-time and store them in the index. This can sometimes be helpful when running cardinality aggregations on high-cardinality and large string fields.

install_remove.asciidoc

Using the murmur3 field

The murmur3 is typically used within a multi-field, so that both the original value and its hash are stored in the index:

PUT my-index-000001
{
  "mappings": {
    "properties": {
      "my_field": {
        "type": "keyword",
        "fields": {
          "hash": {
            "type": "murmur3"
          }
        }
      }
    }
  }
}

Such a mapping would allow to refer to my_field.hash in order to get hashes of the values of the my_field field. This is only useful in order to run cardinality aggregations:

# Example documents
PUT my-index-000001/_doc/1
{
  "my_field": "This is a document"
}

PUT my-index-000001/_doc/2
{
  "my_field": "This is another document"
}

GET my-index-000001/_search
{
  "aggs": {
    "my_field_cardinality": {
      "cardinality": {
        "field": "my_field.hash" (1)
      }
    }
  }
}

1. Counting unique values on the my_field.hash field

Running a cardinality aggregation on the my_field field directly would yield the same result, however using my_field.hash instead might result in a speed-up if the field has a high-cardinality. On the other hand, it is discouraged to use the murmur3 field on numeric fields and string fields that are not almost unique as the use of a murmur3 field is unlikely to bring significant speed-ups, while increasing the amount of disk space required to store the index.