[MS-973] Fix sorting logic in MatchResultSet to handle equal confidence scores

[meladRaouf]

JIRA ticket
Will be released in: 2025.2.0

Root cause analysis (for bugfixes only)

When two subjects have the same confidence score during matching, only one is returned in the identification response. Please refer to the ticket for more details.

First known affected version: All recent versions

• Attached two videos one for the problem and the fix

Notable changes

• Improve sorting logic in MatchResultSet to handle equal confidence scores by subject ID

Testing guidance

• Use the project cGKXBnPkLGKtK0V2TUxo, which contains many duplicate ROCv3 and simatcher subjects, and verify that subjects with the same confidence score are all included in the identification results.

Additional work checklist

• Effect on other features and security has been considered
• Design document marked as "In development" (if applicable)
• External (Gitbook) and internal (Confluence) Documentation is up to date (or ticket created)
• Test cases in Testiny are up to date (or ticket created)
• Other teams notified about the changes (if applicable)

no.duplicates.mp4

allow.duplicates.mp4

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
100.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[BurningAXE]

Hm, am I reading this wrong or if we first add a high confidence score (say 99), then we would not add any lower results and return a list of just 1 result? Is this the behaviour that we want?

[BurningAXE]

Ah, lowestConfidence is updated only when the set is full. That's confusing...

[alex-vt]

It looks like if confidence and subjectId are now a composite comparison key, so should work as intended.

[BurningAXE]

Here we are comparing confidence only. The confusion comes from the fact that lowestConfidence is initialized with 0 and updated only when the set is full.

[meladRaouf]

This is a classic LeetCode problem 🥇. If we only need the top N elements, fully sorting the entire list is unnecessary and can be avoided to enhance performance.

[alex-vt]

It looks like if confidence and subjectId are now a composite comparison key, so should work as intended.