Aggregations in Apache Druid can be used during ingestion to summarize data before it enters Druid, and at query time to summarize result data.
Exact Aggregations
Count Aggregator
Computes the count of Druid rows matching the filters:
{"type": "count", "name": "count"}
| Property | Description | Required |
|---|
type | Must be “count” | Yes |
name | Output name | Yes |
The count aggregator counts Druid rows, not raw ingested events. With rollup enabled, these may differ. To count raw events, include a count aggregator at ingestion time and a longSum aggregator at query time.
Sum Aggregators
longSum
Computes the sum of values as a 64-bit signed integer:
{"type": "longSum", "name": "sumLong", "fieldName": "aLong"}
doubleSum
Computes the sum as a 64-bit floating point value:
{"type": "doubleSum", "name": "sumDouble", "fieldName": "aDouble"}
floatSum
Computes the sum as a 32-bit floating point value:
{"type": "floatSum", "name": "sumFloat", "fieldName": "aFloat"}
| Property | Description | Required |
|---|
type | ”longSum”, “doubleSum”, or “floatSum” | Yes |
name | Output name | Yes |
fieldName | Input column name | No* |
expression | Inline expression | No* |
fieldName or expression
Min and Max Aggregators
doubleMin / doubleMax
{"type": "doubleMin", "name": "minDouble", "fieldName": "aDouble"}
{"type": "doubleMax", "name": "maxDouble", "fieldName": "aDouble"}
floatMin / floatMax
{"type": "floatMin", "name": "minFloat", "fieldName": "aFloat"}
{"type": "floatMax", "name": "maxFloat", "fieldName": "aFloat"}
longMin / longMax
{"type": "longMin", "name": "minLong", "fieldName": "aLong"}
{"type": "longMax", "name": "maxLong", "fieldName": "aLong"}
| Property | Description | Required |
|---|
type | Min/max type (e.g., “doubleMin”) | Yes |
name | Output name | Yes |
fieldName | Input column name | No* |
expression | Inline expression | No* |
fieldName or expression
doubleMean Aggregator
Computes the arithmetic mean as a 64-bit float:
{"type": "doubleMean", "name": "aMean", "fieldName": "aDouble"}
First and Last Aggregators
Return values corresponding to earliest/latest time column values.
With rollup enabled at ingestion, these return rolled-up values, not original raw data values.
Numeric First/Last
doubleFirst / doubleLast:
{
"type": "doubleFirst",
"name": "firstDouble",
"fieldName": "aDouble"
}
{
"type": "floatLast",
"name": "lastFloat",
"fieldName": "aFloat",
"timeColumn": "longTime"
}
{
"type": "longFirst",
"name": "firstLong",
"fieldName": "aLong"
}
| Property | Description | Required |
|---|
type | First/last type | Yes |
name | Output name | Yes |
fieldName | Input column | Yes |
timeColumn | Time column (LONG type) | No (defaults to __time) |
String First/Last
stringFirst:
{
"type": "stringFirst",
"name": "firstString",
"fieldName": "aString",
"maxStringBytes": 2048,
"timeColumn": "longTime"
}
{
"type": "stringLast",
"name": "lastString",
"fieldName": "aString"
}
| Property | Description | Required |
|---|
type | ”stringFirst” or “stringLast” | Yes |
name | Output name | Yes |
fieldName | Input column | Yes |
timeColumn | Time column (LONG type) | No (defaults to __time) |
maxStringBytes | Max string size to accumulate | No (defaults to 1024) |
ANY Aggregators
Return any encountered value (including null). Query time only.
Numeric ANY
doubleAny:
{"type": "doubleAny", "name": "anyDouble", "fieldName": "aDouble"}
{"type": "floatAny", "name": "anyFloat", "fieldName": "aFloat"}
{"type": "longAny", "name": "anyLong", "fieldName": "aLong"}
stringAny
{
"type": "stringAny",
"name": "anyString",
"fieldName": "aString",
"maxStringBytes": 2048,
"aggregateMultipleValues": true
}
| Property | Description | Required |
|---|
type | ”stringAny” | Yes |
name | Output name | Yes |
fieldName | Input column | Yes |
maxStringBytes | Max string size | No (defaults to 1024) |
aggregateMultipleValues | If true, returns stringified array for multi-value dims | No (defaults to true) |
Approximate Aggregations
Count Distinct
DataSketches Theta Sketch
Provides distinct count estimates with set operations (union, intersection, difference):
DataSketches HLL Sketch
HyperLogLog-based distinct counting. More space-efficient than Theta but no set operations:
Legacy: Cardinality and hyperUnique
For new use cases, use DataSketches Theta or HLL instead. Legacy aggregators kept for backwards compatibility.
Histograms and Quantiles
DataSketches Quantiles Sketch (Recommended)
Provides quantile estimates and histogram approximations with formal error bounds:
Moments Sketch (Experimental)
Optimized for merge speed but accuracy is distribution-dependent:
Fixed Buckets Histogram
Simple histogram with fixed range and bucket count:
Approximate Histogram (Deprecated)
Deprecated due to accuracy issues. Use DataSketches Quantiles instead.
Expression Aggregations
Expression Aggregator
Custom aggregations using Druid expressions (query time only):
{
"type": "expression",
"name": "expression_sum",
"fields": ["column_a"],
"accumulatorIdentifier": "__acc",
"initialValue": "0",
"fold": "__acc + column_a",
"combine": "__acc + expression_sum"
}
| Property | Description | Required |
|---|
type | Must be “expression” | Yes |
name | Output name | Yes |
fields | Input columns | Yes |
accumulatorIdentifier | Variable for accumulator | No (defaults to __acc) |
fold | Expression to accumulate values | Yes |
combine | Expression to merge fold results | No (defaults to fold) |
compare | Comparator expression (inputs: o1, o2) | No |
finalize | Final transformation (input: o) | No |
initialValue | Initial accumulator value | Yes |
initialCombineValue | Initial combine value | No (defaults to initialValue) |
isNullUnlessAggregated | Return null if no rows processed | No (defaults to true) |
shouldAggregateNullInputs | Process null inputs in fold | No (defaults to true) |
shouldCombineAggregateNullInputs | Process null inputs in combine | No (defaults to shouldAggregateNullInputs) |
maxSizeBytes | Max size for variable-sized outputs | No (defaults to 8192) |
Example: Count Aggregator
{
"type": "expression",
"name": "expression_count",
"fields": [],
"initialValue": "0",
"fold": "__acc + 1",
"combine": "__acc + expression_count"
}
Example: Distinct Array with Ordering
{
"type": "expression",
"name": "expression_array_agg_distinct",
"fields": ["column_a"],
"initialValue": "[]",
"fold": "array_set_add(__acc, column_a)",
"combine": "array_set_add_all(__acc, expression_array_agg_distinct)",
"compare": "if(array_length(o1) > array_length(o2), 1, if(array_length(o1) == array_length(o2), 0, -1))"
}
JavaScript Aggregator
Custom JavaScript functions for aggregation:
{
"type": "javascript",
"name": "sum(log(x)*y) + 10",
"fieldNames": ["x", "y"],
"fnAggregate": "function(current, a, b) { return current + (Math.log(a) * b); }",
"fnCombine": "function(partialA, partialB) { return partialA + partialB; }",
"fnReset": "function() { return 10; }"
}
Miscellaneous Aggregations
Filtered Aggregator
Wraps any aggregator to only aggregate matching rows:
{
"type": "filtered",
"filter": {
"type": "selector",
"dimension": "someColumn",
"value": "abcdef"
},
"aggregator": {
"type": "longSum",
"name": "sumLong",
"fieldName": "aLong"
}
}
If you only need filtered results, put the filter on the query itself for better performance.
Grouping Aggregator
For GroupBy queries with subtotals - returns a number indicating which dimensions are in the sub-grouping:
{
"type": "grouping",
"name": "someGrouping",
"groupings": ["dim1", "dim2"]
}
[["dim1", "dim2"], ["dim1"], ["dim2"], []]:
| Subtotal | Output | Binary |
|---|
["dim1", "dim2"] | 0 | (00) |
["dim1"] | 1 | (01) |
["dim2"] | 2 | (10) |
[] | 3 | (11) |