Elasticsearch aggregation float type losing precision - elasticsearch

Question

If you use Elasticsearch 5.5 with Dynamic field mapping
and use double values. These values are getting the float type when I check in the mappings. When you are using an aggregation than the key in the buckets will be losing precision. the Value 0.62 would be something like 0.6200000047683716.
Code fragment
"aggregations": {
"float_numbers": {
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0,
"buckets": [
{
"key": 0.6200000047683716,
"doc_count": 1
}
]
}
}
Here is the same issue described.
link
I am posting this issue because I found an appropriate solution which I not yet have seen but it helped me a lot.
The solution is to make the float a double. This can be achieved with Dynamic templates.
dynamic templates
dynamic field mapping
Example solution:
Add dynamic_templates in index there are no items yet.
PUT term-test
{
"mappings": {
"demo_typ": {
"dynamic_templates": [
{
"all_to_double": {
"match_mapping_type": "double",
"mapping": {
"type": "double"
}
}
}
]
}
}
}
Add data
POST term-test/demo_typ
{
"numeric_field": 0.62,
"long_filed": 44
}
Check mapping
GET term-test/_mapping
Do aggregation
GET term-test/_search
{
"query": {
"match_all": {}
},
"aggs": {
"float_numbers": {
"terms": {
"field": "numeric_field"
}
}
}
}
In the Java Api you can do the following
1: First create the index
elasticClient.admin()
.indices()
.prepareCreate(indexName)
.execute()
.actionGet();
2: Update the mapping
JSON
{
"dynamic_templates": [
{
"all_to_double": {
"match_mapping_type": "double",
"mapping": {
"type": "double"
}
}
}
]
}
Json to XContentBuilder I got the code from link
public XContentBuilder getXContentBuilderFromJson(final String json) {
try {
Map<String, Object> map = new ObjectMapper().readValue(json, new TypeReference<Map<String, Object>>() {});
return XContentFactory.jsonBuilder().map(map);
} catch (IOException e) {
e.printStackTrace();
return null;
}
}
Update mapping
elasticClient.admin().indices()
.preparePutMapping(indexName)
.setType(yourType)
.setSource(getXContentBuilderFromJson(json))
.execute()
.actionGet();
3: Insert data

Answer 1

Numbers lose precision. This is because of how floating-point numbers work: 9.62 can't be expressed as a * 2 ^ b so neither doubles nor floats can represent it accurately.
Because floats and doubles cannot accurately represent a value, it is generally a bad idea to run terms aggregations on them.
As a workaround you can do Math.round after you did the aggregation