Post aggregation example query for druid in json - druid

Question

I am trying to use post aggregation. I have used aggregation to count the number of rows which match the given filter. Following is the post aggregation query:
{
"queryType": "groupBy",
"dataSource": "datasrc1",
"intervals": ["2020-09-16T21:15/2020-09-16T22:30"],
"pagingSpec":{ "threshold":100},
"dimensions": ["city", "zip_code", "country"],
"filter": {
"fields": [
{
"type": "selector",
"dimension": "bankId",
"value": "<bank id>"
}
]
},
"granularity": "all",
"aggregations": [
{ "type": "count", "name": "row"}
],
"postAggregations": [
{ "type": "arithmetic",
"name": "sum_rows",
"fn": "+",
"fields": [
{ "type": "fieldAccess", "fieldName": "row" }
]
}
]
}
If I remove the post aggregation part, it returns me result like:
[ {
"version" : "v1",
"timestamp" : "2020-09-16T21:15:00.000Z",
"event" : {
"city": "Sunnyvale",
"zip_code": "94085",
"country": "US",
"row" : 1
}
}, {
"version" : "v1",
"timestamp" : "2020-09-16T21:15:00.000Z",
"event" : {
"city": "Sunnyvale",
"zip_code": "94080",
"country": "US",
"row" : 1
}
}
If I add the post aggregations part, I get parser exception:
{
"error" : "Unknown exception",
"errorMessage" : "Instantiation of [simple type, class io.druid.query.aggregation.post.ArithmeticPostAggregator] value failed: Illegal number of fields[
%s], must be > 1 (through reference chain: java.util.ArrayList[0])",
"errorClass" : "com.fasterxml.jackson.databind.JsonMappingException",
"host" : null
}
I want to add all the rows (column 'row') in the response we are getting for aggregation query; and put the output in "sum_rows".
I don't understand what I am missing in post_aggregations. Any help is appreciated.

Answer 1

Confess that I spend most of my time in the SQL API not in the native API (!!) but I think your issue is that you're only supplying one field to your post aggregator. See these examples:
https://druid.apache.org/docs/latest/querying/post-aggregations.html#example-usage
If you need sum of rows, perhaps you need a normal aggregator to sum the row count?

Answer 2

The error message says the ArithmeticPostAggregator requires 2 arguments; the example code has only one. There's an example of this post aggregator at the bottom of this answer.
However...the example query doesn't have multiple numeric aggregations to perform arithmetic post-aggregation against. Maybe the goal is to "combine" the two output rows into one?
...To change the two-row result into only one with the total row count (for all database records matching the query filter and interval), removing zip_code from the dimension list would be one way.
Removing zip_code from dimensions would produce one result like this:
[
{
"version" : "v1",
"timestamp" : "2020-09-16T21:15:00.000Z",
"event" : {
"city": "Sunnyvale",
"country": "US",
"row" : 2
}
]
As you can see, by submitting a groupBy query with aggregations, Druid will do this aggregation for you dynamically (based on the dimension values in the database at the time the query is run) without needing post aggregations.
Example arithmetic post aggregator:
{
"type": "arithmetic",
"name": "my_output_sum",
"fn": "+",
"fields": [
{"fieldName": "input_addend_1", "type":"fieldAccess"},
{"fieldName": "input_addend_2", "type":"fieldAccess"}
]
}