This function is a filter query that runs in two phases.

  1. Runs a query to determine a set of IDs (specified using the field parameter), for which there exists an event with that field ID which satisfy all the where clauses. Each where clause can be satisfied by distinct events (but they must all have the same ID).

  2. Runs as a filter function that lets all events that have one of the determined IDs pass through. In the secondary run, the events need only match the ID, not any of the where clauses; unless prefilter=true is set.

field[a]Array of stringsrequired  Specifies which field in the event (log line) is the join key identifier.
prefilterbooleanoptional[b]false Only pass through values matching at least one of the where clauses.
where[Filter]required  The subquery to execute producing the values to join with.

[a] The argument name field can be omitted.

[b] Optional parameters use their default value unless explicitly set

Omitted Argument Names

The argument name for field can be omitted; the following forms of this function are equivalent:




These examples show basic structure only; full examples are provided below.

selfJoinFilter() is probabilistic and the result can contain false positives.

matches false positive rate number of false positives
1000 0.00000% 0.0
10000 0.00029% 0.0
20000 0.00224% 0.4
25000 0.00434% 1.1
50000 0.03289% 16.4

If, for example, the where clauses (along with any preceding filtering) limits the matching IDs to 25,000 elements, then out of those 1.1 will be false positives on average.

When passed the additional argument prefilter=true, the resulting output will only contain those log lines that match one of the where clauses. With prefilter set to false by default, all log lines with a join key for which there exists events that satisfy the where clauses will be passed through.

This function does two passes over the data and can therefore not be used in a live query unless in combination with beta:repeating().


If multiple fields are specified in the field parameter, they must all exist in an event, for it to be valid for selfJoinFilter().

You have emails logged with one event for each header. Find all attachments for emails sent from Peter to Paul, by first finding all the email_ids that correspond to mails from Peter to Paul, subsequently find all log messages with one of those email_ids that also has an attachment.

selfJoinFilter(field=email_id, where=[{ from=peter }, {to=paul}])
| attachment=*