Summary

Jira creates tokens when searching for text fields.
For a case like A-B-C Jira can create three different tokens or Jira can keep these three words as only one token. The same behavior happens for the pattern A-B, where sometimes creates two separated tokens and other times creates only one token.

Since Jira will try to match the words used on the JQL with each token, and Jira can create tokens differently in each case, that's the reason why sometimes a JQL works for one case but not for another case.

Update (April 26, 2021)

Any custom Text Type custom field also is affected.

Update (April 3, 2020)

Summary search is affected as well.

Searching for !~ is also impacted:

• Issues with summary: something, test - test, test;

  JQL	Result
  summary !~ "test - test"	returns all issues
  summary !~ "test"

Environment

• JIRA Cloud

Steps to Reproduce

1. Create New project
2. Create test tickets
   • Description: release-abc
3. Go to issue navigator: https://<EXAMPLE>.atlassian.net/issues/?jql=
4. Conduct the following test using Advance JQL

JQL	Result
description ~ "release-abc"
description ~ "-abc
description ~ "abc"
description ~ "release"
description ~ "release-"

 = Issue not found
 = Issue found

Expected Results

• Find issues with description matching the JQL

Actual Results

• Two scenarios where results did not populate correctly

Notes

• Quick Search also displays same results:
  • https://<EXAMPLE>.atlassian.net/secure/QuickSearch.jspa?searchString=release-abc > 

Workaround

• Using specific text search might work for some cases:
  • summary ~ "\"test - test\""
  • summary !~ "\"test - test \""
    It does seem inconsistent sometimes even with an exact search.

• Also changing the Indexing Language under System Settings to English - Moderate Stemming seem to help.