关于fielddata疑问

es版本：5.4.0
看官网文档提到fielddata用于分析字段的聚合，查询fielddata内存使用情况输出如下所示，为什么url.keyword字段使用了fielddata内存区，url.keyword不应该是not_analyzed的字段，聚合时使用的是doc values方式吗？
"total": {
      "fielddata": {
        "memory_size_in_bytes": 1739416,
        "evictions": 0,
        "fields": {
          "source_node.transport_address": {
            "memory_size_in_bytes": 0
          },
          "source_node.name": {
            "memory_size_in_bytes": 0
          },
          "cluster_uuid": {
            "memory_size_in_bytes": 0
          },
          "method.keyword": {
            "memory_size_in_bytes": 62928
          },
          "source_node.uuid": {
            "memory_size_in_bytes": 0
          },
          "logstash_stats.logstash.uuid": {
            "memory_size_in_bytes": 0
          },
          "shard.index": {
            "memory_size_in_bytes": 0
          },
          "kibana_stats.kibana.status": {
            "memory_size_in_bytes": 0
          },
          "url.keyword": {
            "memory_size_in_bytes": 1676488
          },
          "kibana_stats.kibana.uuid": {
            "memory_size_in_bytes": 0
          },
          "_parent": {
            "memory_size_in_bytes": 0
          },
          "shard.state": {
            "memory_size_in_bytes": 0
          }
        }
      }
    }
  }