Join search result rows with other search result rows in the same result set, based on one or more fields that you specify.
selfjoin [<selfjoin-options>...] <field-list>
- Syntax: <field>...
- Description: The field or list of fields to join on.
- Syntax: overwrite=<bool> | max=<int> | keepsingle=<bool>
- Description: Options that control the search result set that is returned. You can specify one or more of these options.
- Syntax: keepsingle=<bool>
- Description: Controls whether or not to retain results that have with a unique value in the join fields. When
keepsingle=truesearch results that have no other results to join with are kept in the output.
- Default: false
- Syntax: max=<int>
- Description: Indicates the maximum number of 'other' results to join with each main result. If
max=0, there is no limit. This argument sets the maximum for the 'other' results. The maximum number of main results is 100,000.
- Default: 1
- Sytnax: overwrite=<bool>
- Description: When
overwrite=true, causes fields from the 'other' results to overwrite fields of the main results. The main results are used as the basis for the join.
- Default: true
Self joins are more commonly used with relational database tables. They are used less commonly with event data.
An example of an events usecase is with events that contain information about processes, where each process has a parent process ID. You can use the
selfjoin command to correlate information about a process with information about the parent process.
See the Extended example.
1: Use a single field to join results
Join the results with itself on the 'id' field.
... | selfjoin id
The following example shows how the
selfjoin command works against a simple set of results.
You can follow along with this example on your own Splunk instance.
This example builds a search incrementally. With each addition to the search, the search is rerun and the impact of the additions are shown in a results table. The values in the
_time field change each time you rerun the search. However, in this example the values in the results table are not changed so that we can focus on how the changes to the search impact the results.
1. Start by creating a simple set of 5 results by using the
| makeresults count=5
There are 5 results created, each with the same timestamp.
2. To keep better track of each result use the
streamstats command to add a field that numbers each result.
| makeresults count=5 | streamstats count as a
a field is added to the results.
3. Additionally, use the
eval command to change the timestamps to be 60 seconds apart. Different timestamps makes this example more realistic.
| makeresults count=5 | streamstats count as a | eval _time = _time + (60*a)
The minute portion of the timestamp is updated.
4. Next use the
eval command to create a field to use as the field to join the results on.
| makeresults count=5 | streamstats count as a | eval _time = _time + (60*a) | eval joiner="x"
The new field is added.
5. Use the
eval command to create some fields with data.
if function is used with a modulo (modulus) operation to add different data to each of the new fields. A modulo operation finds the remainder after the division of one number by another number:
eval bcommand processes each result and performs a modulo operation. If the remainder of
a/2is 0, put “something” into the field “b”, otherwise put "nada" into field “b”.
eval ccommand processes each result and performs a modulo operation. If the remainder
a/2is 1, put “something else” into the field “c”, otherwise put nothing (NULL) into field “c”.
| makeresults count=5 | streamstats count as a | eval _time = _time + (60*a) | eval joiner="x" | eval b = if(a%2==0,"something","nada"), c = if(a%2==1,"somethingelse",null())
The new fields are added and the fields are arranged in alphabetical order by field name, except for the
6. Use the
selfjoin command to join the results on the
| makeresults count=5 | streamstats count as a | eval _time = _time + (60*a) | eval joiner="x" | eval b = if(a%2==0,"something","nada"), c = if(a%2==1,"somethingelse",null()) | selfjoin joiner
The results are joined.
7. To understand how the
selfjoin command joins the results together, remove the
| selfjoin joiner portion of the search. Then modify the search to append the values from the
a field to the values in the
| makeresults count=5 | streamstats count as a | eval _time = _time + (60*a) | eval joiner="x" | eval b = if(a%2==0,"something"+a,"nada"+a), c = if(a%2==1,"somethingelse"+a,null())
The results now have the row number appended to the values in the
8. Now add the
selfjoin command back into the search.
| makeresults count=5 | streamstats count as a | eval _time = _time + (60*a) | eval joiner="x" | eval b = if(a%2==0,"something"+a,"nada"+a), c = if(a%2==1,"somethingelse"+a,null()) | selfjoin joiner
The results of the self join.
If there are values for a field in both rows, the last result row, based on the
_time value, takes precedence. The joins performed are shown in the following table.
|1||Row 1 is joined with row 2 and returned as row 2.||In field |
|2||Row 2 is joined with row 3 and returned as row 3.||Since row 3 contains values for both field |
|3||Row 3 is joined with row 4 and returned as row 4.||In field |
|4||Row 4 is joined with row 5 and returned as row 5.||Since row 5 contains values for both field |
|5||Row 5 has no other row to join with.||No additional results are returned.|
(Thanks to Splunk user Alacercogitatus for helping with this example.)
Have questions? Visit Splunk Answers and see what questions and answers the Splunk community has using the selfjoin command.
This documentation applies to the following versions of Splunk Cloud™: 6.5.0, 6.5.1, 6.6.0, 6.5.1612, 6.6.1, 6.6.3, 7.0.0