翻译强力驱动I would like to use the stats or extended_stats aggregation on the _score field but can't find any examples of this being done (i.e., seems like you can only use aggregations with actual document fields).
Is it possible to request aggregations on calculated "metadata" fields for each hit in an ElasticSearch query response (e.g., _score, _type, _shard, etc.)?
I'm assuming the answer is 'no' since fields like _score aren't indexed...
Note: The original answer is now outdated in terms of the latest version of Elasticsearch. The equivalent script using Groovy scripting would be:
{ ..., "aggregations" : { "grades_stats" : { "stats" : { "script" : "_score" } } } }In order to make this work, you will need to enable dynamic scripting or, even better, store a file-based script and execute it by name (for added security by not enabling dynamic scripting)!
You can use a script and refer to the score using doc.score. More details are available in ElasticSearch's scripting documentation.
A sample stats aggregation could be:
{ ..., "aggregations" : { "grades_stats" : { "stats" : { "script" : "doc.score" } } } }And the results would look like:
"aggregations": { "grades_stats": { "count": 165, "min": 0.46667441725730896, "max": 3.1525731086730957, "avg": 0.8296855776598959, "sum": 136.89812031388283 } }A histogram may also be a useful aggregation:
"aggs": { "grades_histogram": { "histogram": { "script": "doc.score * 10", "interval": 3 } } }Histogram results:
"aggregations": { "grades_histogram": { "buckets": [ { "key": 3, "doc_count": 15 }, { "key": 6, "doc_count": 103 }, { "key": 9, "doc_count": 46 }, { "key": 30, "doc_count": 1 } ] } }doc.score doesn't seem to work anymore. Using _score seems to work perfectly.
Example:
{ ..., "aggregations" : { "grades_stats" : { "stats" : { "script" : "_score" } } } }