Since the most common use-case is to count the distinct values of a field, the default behavior is to use count() to aggregate the events. The simplest use-case of groupBy() is, therefore, to count the occurrence of the distinct values of a field:

groupBy(status_code)

Given input:

The output is:

The task is to only find the unique values, groupBy() can be provided with an empty list as the aggregate function, implying that nothing will be aggregated:

groupBy(status_code, function=[])

The output is:

Similarly, groupBy() can be used to find the unique combinations of field values present in the data:

groupBy([status_code, ip], function=[])

The output is:

A usage of the function such as the one above is also very useful for query-based parameters in dashboards, see Query Parameter for more explanations.

As an example of how to use this query function, suppose you have a LogScale repository that is ingesting data from one or more web servers. Now suppose you want to get a list of HTTP status codes, the results of users accessing the server. For instance, HTTP code 404 for Not Found — for web pages not found. You could get a list of all of the HTTP codes returned and a count of each by executing a query that counts the different http status codes, like this:

groupBy(field=status_code, function=count())
| sort(statuscode, order=asc)

In addition to the groupBy(), the results here are piped to the sort() function to sort the results from the lowest HTTP code to the highest.

With the groupBy() function, you can have LogScale group by more than one field. You would just give the fields in an array, a comma-separated list within square-brackets. It would look something like this query in which events are grouped on HTTP method and status code:

groupBy(field=[method, statuscode], function=count())
| sort([method, statuscode], order=asc)

Will produce these results:

Although this query works, you may also want to show the total for each method, not just the total for each status code of each method. To do that, you'll have to adjust the query to look like this:

[groupby(method, function=[count(as=method_total), groupby(statuscode, function=count(as=method_status_count))])]

In this query, a groupBy() is nested within another groupBy(). The results will look like this:

These results still might not be as tidy as you like. You might do better with separate queries for each method or each status code — assuming you're interested in only a few specific ones. Another possibility might be to use an API to be able to assemble the results in a manner that you prefer.

Instead of a single function, additional query functions can be also applied, allowing for calculations on the count that is obtained from groupBy(). The following query contains an embedded expression within groupBy(), as a demonstration of formatting/calculating the value within a groupBy() expression:

groupBy(host, function=[{count()
| esp:=_count/300}])

The query generates a count() of the number of hosts in the aggregation, which creates a field _count which is then divided by 300 and placed into the field esp.