Aggregation commands

fieldsSummary

The fieldsSummary command calculates the cardinality of field values that the specified fields have.

Syntax

fieldsSummary field, … [, topValues] [, extrapolateSamples]

Parameters

ParameterTypeDescriptionRequired

field

field identifier

A field identifier.

required

topValues

positive long

The number of top N values to be returned.

optional

extrapolateSamples

boolean

Flag indicating if the cardinality shall be multiplied with a possible sampling rate.

optional

Basic example

Example: Simple fields summary

The following example shows the cardinality of values for the field host.

data record(host = "host-a"),
record(host = "host-a"),
record(host = "host-b")
| fieldsSummary host

Query result:

field
rawCount
count
values
host
3
3
[value: host-a count: 2, value: host-b count: 1]

Practical example

Example: Log count by host

In this example, the query fetches logs with a sampling ratio of 10,000 and calculates the cardinality of values for the dt.entity.host field, providing the 10 top values encountered when extrapolating the value count by the sampling ratio.

fetch logs, samplingRatio: 10000
| fieldsSummary dt.entity.host, topValues: 10, extrapolateSamples: true

makeTimeseries

Creates timeseries from the data in the stream. The makeTimeseries command provides a convenient way to chart raw non-metric data (such as events or logs) over time.

Syntax

makeTimeseries [by: { [expression, …] }] [, interval] [, bins] [, from] [, to] [, timeframe] [, time ,] [, spread ,] [, nonempty ,] aggregation, …

Parameters

ParameterTypeDescriptionRequired

aggregation

aggregation function

The aggregation function that should be used to create the series.
Input values in the aggregation function need to be numeric.

required

time

expression

The expression that provides the timestamp for a record, which will be assigned to a series bucket. If the current data does not contain a field timestamp the parameter needs to be specified, referencing the time field. The default value is timestamp.

optional

from

timestamp, duration

The start timestamp of the series. The actual series timeframe will be adjusted by the series alignment.

optional

to

timestamp, duration

The end timestamp of the series. The actual series timeframe will be adjusted by the series alignment.

optional

timeframe

timeframe

The timeframe of the series.
The actual series timeframe will be adjusted by the series alignment. The default value is query timeframe.

optional

bins

positive integer

The number of buckets that should be created for the series.
Must not be specified if interval is specified. The default value is 120.

optional

interval

positive duration

The length of a bucket in the series. Must not be specified if bins is specified. Automatically calculated based on the number of bins.

optional

by

list of expressions

The expressions the series should be split by.

optional

spread

timeframe

A timeframe expression that provides the timeframe for the bucket calculation of the values in the series. Spread can only be used instead of time and you can only use it with the count or countIf functions.

optional

nonempty

boolean

Produces empty series when there is no data.

optional

Basic example

Example: Count records

The following example counts the number of records per time interval.

data record(timestamp = toTimestamp("2019-08-01T09:30:00.000-0400")),
record(timestamp = toTimestamp("2019-08-01T09:31:00.000-0400")),
record(timestamp = toTimestamp("2019-08-01T09:31:30.000-0400")),
record(timestamp = toTimestamp("2019-08-01T09:32:00.000-0400"))
| makeTimeseries count(),
from: toTimestamp("2019-08-01T09:30:00.000-0400"),
to: toTimestamp("2019-08-01T09:33:00.000-0400")

Query result:

timeframe
interval
count()
start: 2019-08-01T13:30:00.000Z
end: 2019-08-01T13:33:00.000Z
1 min
[1, 2, 1]

Practical examples

Example 1: Error logs for host

This example shows error logs for a host over time.

fetch logs
| filter dt.entity.host == "HOST-02A6022668FCA28E" and loglevel == "ERROR"
| makeTimeseries count()
Example 2: Buy activity for account

The following example shows the buy activity for a specific account for the last 24 hours. The interval parameter sets the time interval to 30 minutes, resulting in 48 bins in the query outcome. A bin with no buy activity has the value 0 because the default parameter is 0. Not setting this parameter results in NULL values for the bins.

fetch bizevents, from: now() - 24h
| filter accountId == 7
| filter in(event.category, array("/broker-service/v1/trade/long/buy", "/v1/trade/buy"))
| makeTimeseries count(default: 0), interval: 30m
Example 3: Transaction statistics

The following advanced example takes sell transactions of the last seven days recorded as business events and charts the total number of transactions and high-volume transactions based on the conditional countIf() function and the maximum price by accountId. The interval parameter defines a fixed chart resolution of 1 day.

fetch bizevents, from: now() - 7d
| filter in(event.type, array("com.easytrade.long-sell", "com.easytrade.quick-sell"))
| makeTimeseries {
count(),
high_volume = countIf(amount >= 100 and amount <= 10000),
max(price)
},
by: { accountId },
interval: 1d

Bins parameter

The interval and bins parameters are exclusive. Both parameters are optional.

If specified, the bins parameter (range: 12–1,500) is used to calculate an equivalent nominal time interval.

The nominal time interval, whether coming from the bins or interval parameter, is adjusted so that:

  • The time interval matches a well-known interval (1 minute, 5 minutes, 10 minutes, 15 minutes, 30 minutes, or 1 hour)
  • The time interval does not exceed the maximum number of elements per series (1,500).

Default value for empty time slots

The makeTimeseries command produces homogenous time series of aggregated data: all series have identical start and end timestamps, time interval and number of elements. If data is missing for a particular time slot, it is filled with null. Specifying a default parameter fills empty time slots with the default parameter value instead of null.

There might be no record processed by the makeTimeseries command. For example, when no relevant data is available for the timeframe. The makeTimeseries command returns an empty result in this case. If the expected result is a chart with 0 values, using nonempty: true in combination with default: 0 produces the desired result.

Example 1
fetch logs
| filter status >= 500
| makeTimeseries count = count(default: 0), interval: 30m

Query result:

timeframeintervalcount

No records

Example 2
fetch logs
| filter status >= 500
| makeTimeseries count = count(default: 0), interval: 30m, nonempty: true

Query result:

timeframeintervalcount

start: 2019-08-01T13:00:00.000Zend: 2019-08-01T15:30:00.000Z

30 min

[0, 0, 0, 0, 0]

Aggregation functions

Nine aggregation functions are available to use with the makeTimeseries command. These functions are: min, max, sum, avg, count, countIf, percentile, countDistinctExact and countDistincApprox. It is also possible to use start and end with makeTimeseries, however, they need to be used together with another aggregation function.

Syntax

These functions have the following syntax:

  • sum(expression [, default] [, rate])
  • avg(expression [, default] [, rate])
  • min(expression [, default] [, rate])
  • max(expression [, default] [, rate])
  • count([default] [, rate])
  • countIf(expression [, default] [, rate])
  • countDistinctExact(expression [, default] [, rate])
  • countDistinctApprox(expression [, precision] [, default] [, rate])
  • start()
  • end()
  • percentile(expression, percentile [, default] [, rate])
Parameters
ParameterTypeDescriptionRequired

expression

expression

The expression the series should be created for.

required

default

number

The default value that should be used to fill gaps/empty bins. The default value is null.

optional

rate

duration

The duration that shall be used to adjust the bin values using the following formula: (binValue / interval) * rate.

optional

percentile

double, long

The nth-percentile, such that approximately n percent of observed measurements fall below this value. Must be between 0 and 100.

required

precision

long

Parameter that sets the precision level of the estimation. Must be between 3 and 16. The default value is 14.

optional

summarize

Groups together records that have the same values for a given field and aggregates them.

Syntax

summarize [field =] aggregation, ... [, by: {[field =] expression, ...}]

Parameters

ParameterTypeDescriptionRequired

expression

array, boolean, counter, double, duration, ip, long, record, string, timeframe, timestamp

An expression to group by.

required

aggregation

array, boolean, counter, double, duration, ip, long, record, string, timeframe, timestamp

An aggregation function

required

Aggregation functions

You can use different aggregation functions with the summarize command. See all available aggregation functions.

Basic examples

Example 1: Sum field

The below example uses the summarize command and the sum aggregation function to sum the field value.

data record(value = 2),
record(value = 3),
record(value = 7),
record(value = 7),
record(value = 1)
| summarize sum(value)

Query result:

sum(value)
20
Example 2: Summarize by category

The following example calculates the sum of the field value and splits the result by the field cat. Furthermore, the resulting aggregation field gets the alias name sum and the grouping field the alias category.

data record(value = 2, cat = "a"),
record(value = 3, cat = "b"),
record(value = 7, cat = "a"),
record(value = 7, cat = "b"),
record(value = 1, cat = "b")
| summarize sum = sum(value),
by: { category = cat }

Query result:

category
sum
a
9
b
11
Example 3: Empty aggregation result

If the summarize command has no input records and you don't use the by clause, the summarize command will still return a single record a as result.

data record(value = 2),
record(value = 3),
record(value = 7),
record(value = 7),
record(value = 1)
| filter value > 7
| summarize count(), sum(value), collectArray(value), takeAny(value)

Query result:

count()
sum(value)
collectArray(value)
takeAny(value)
0
null
null
null
Example 4: Group by field doesn't exist for every record

If the grouped field does not exist for every record, the summarize command adds a null group to the result.

data record(value = 2),
record(value = 3, category = "b"),
record(value = 7, category = "a"),
record(value = 7, category = "b"),
record(value = 1)
| summarize sum(value),
by: { category }

Query result:

category
sum(value)
a
7
b
10
null
3
Example 5: Using summarize instead of join

You can use the summarize command instead of the join command when joining on the same table. As input for the summarize command, there are two types of records. In this example, there are records with fields key and value and records with fields key and amount. The key field is set as the by parameter. The other fields are added to the aggregation list. If a record type has duplicate key values, use the aggregation function collectArray for its fields. Otherwise, use the takeAny aggregation function for its record type fields. As the last step, you need to expand the fields that use the collectArray function. Use the summarize command instead of the join to avoid fetching the same table twice.

data record(key = "a", value = 1),
record(key = "b", value = 2),
record(key = "c", value = 4),
record(key = "b", amount = 10),
record(key = "c", amount = 20),
record(key = "c", amount = 40),
record(key = "d", amount = 50)
| summarize {
value = takeAny(value),
amount = arrayRemoveNulls(collectArray(amount))
},
by: { key }
| expand amount

Query result:

key
value
amount
b
2
10
c
4
20
c
4
40
d
null
50
Example 6: Element-wise aggregation

The following example uses 'summarize' with an iterative expression in the 'sum' aggregation function to calculate the element-wise sum of arrays in the input records.

data record(a = array(2, 2)),
record(a = array(7, 1))
| summarize sum(a[])

Query result:

sum(a[])
[9, 3]

Practical example

Example: Count selected log levels

The following example counts selected log levels and groups them by dt.entity.host and dy.entity.process_group.

fetch logs
| summarize {
errors = countIf(loglevel == "ERROR"),
warnings = countIf(loglevel == "WARN"),
severe = countIf(loglevel == "DEBUG")
},
by: {
dt.entity.host,
dt.entity.process_group
}