Splunk Search

Alternative to streamstats+last

duesser
Path Finder

I have this query, where I want to build a dataset from a variable and its 4 previous values. I can solve this like so:

 

 

 

| makeresults
| eval id=split("a,b,c,d,e,f,g",",")
| eval a=split("1,2,3,4,5,6,7",",")
| eval temp=mvzip(id,a,"|")
| mvexpand temp
| rex field=temp "(?P<id>[^|]+)\|(?P<a>[^|]+)"
| fields - temp
| streamstats current=false last(a) AS a_lag1
| streamstats current=false last(a_lag1) AS a_lag2
| streamstats current=false last(a_lag2) AS a_lag3
| streamstats current=false last(a_lag3) AS a_lag4
| where isnotnull(a_lag4)
| table id a*

 

 

 

However, if I want to extend this to say 100 previous values, this code would become convoluted and slow. I imagine there must be a better way to accomplish this goal, however my research has not produced any alternative.

Any ideas are appreciated.

0 Karma
1 Solution

ITWhisperer
SplunkTrust
SplunkTrust

Use autoregress something like this

| makeresults
| eval a=mvrange(1,102)
| mvexpand a
| autoregress a p=1-100

View solution in original post

ITWhisperer
SplunkTrust
SplunkTrust

Use autoregress something like this

| makeresults
| eval a=mvrange(1,102)
| mvexpand a
| autoregress a p=1-100

PickleRick
SplunkTrust
SplunkTrust

You mean something like

| streamstats <optionally current=f> window=10 list(myfield) <optionally BY another_field>
Get Updates on the Splunk Community!

Best Strategies to Optimize Observability Costs

 Join us on Tuesday, May 6, 2025, at 11 AM PDT / 2 PM EDT for an insightful session on optimizing ...

Fueling your curiosity with new Splunk ILT and eLearning courses

At Splunk Education, we’re driven by curiosity—both ours and yours! That’s why we’re committed to delivering ...

Splunk AI Assistant for SPL 1.1.0 | Now Personalized to Your Environment for Greater ...

Splunk AI Assistant for SPL has transformed how users interact with Splunk, making it easier than ever to ...