Splunk Search

Magical sort order

yuanliu
SplunkTrust
SplunkTrust

I was surprised by this result: In a field starting with a value that can be interpreted as an integer, groupby treats it lexically, but sort treats it numerically.  How does sort determine the intention?  Is there a syntax to force lexical sort?

To illustrate, consider the following:

 

| makeresults
| eval i = mvrange(-3, 4)
| mvexpand i
| eval i = printf("%+d", i) . "x"
| stats count by i

 

 

 

  • As is (groupby only)
    icount
    +0x1
    +1x1
    +2x1
    +3x1
    -1x1
    -2x1
    -3x1
  • Add |sort i
    icount
    -3x1
    -2x1
    -1x1
    +0x1
    +1x1
    +2x1
    +3x1

In my use case, numeric sort is desired. (That was how I "discovered" this.)  Just curious about mechanism.

Labels (1)
Tags (1)
0 Karma
1 Solution

richgalloway
SplunkTrust
SplunkTrust

To see how sort determines how to sort the results, see https://docs.splunk.com/Documentation/Splunk/8.2.4/SearchReference/Sort#Usage . You can override the default by specifying a Sort Field Option.  See https://docs.splunk.com/Documentation/Splunk/8.2.4/SearchReference/Sort#Sort_field_options

Since the manual entry for stats is silent on the subject, I presume it uses lexicographical order.

---
If this reply helps you, Karma would be appreciated.

View solution in original post

richgalloway
SplunkTrust
SplunkTrust

To see how sort determines how to sort the results, see https://docs.splunk.com/Documentation/Splunk/8.2.4/SearchReference/Sort#Usage . You can override the default by specifying a Sort Field Option.  See https://docs.splunk.com/Documentation/Splunk/8.2.4/SearchReference/Sort#Sort_field_options

Since the manual entry for stats is silent on the subject, I presume it uses lexicographical order.

---
If this reply helps you, Karma would be appreciated.

yuanliu
SplunkTrust
SplunkTrust

Thanks for the pointer!  In short, to force lexicographical order, | sort str(i). (I had used sort ip() but didn't know str() was also a directive.)

0 Karma
Get Updates on the Splunk Community!

Data Management Digest – December 2025

Welcome to the December edition of Data Management Digest! As we continue our journey of data innovation, the ...

Index This | What is broken 80% of the time by February?

December 2025 Edition   Hayyy Splunk Education Enthusiasts and the Eternally Curious!    We’re back with this ...

Unlock Faster Time-to-Value on Edge and Ingest Processor with New SPL2 Pipeline ...

Hello Splunk Community,   We're thrilled to share an exciting update that will help you manage your data more ...