[Deepya1502]

Summary

• Calculate relative unique-leaf paths from schema ancestors when a list has no absolute schema path, instead of throwing a NullPointerException.
• Add an anydata payload-list fixture with nested unique "details/name" and test both duplicate and distinct values.

This follows up the separate unique null-schema-path review on #50; it does not change anydata strictness.

Validation

• mvn -q -pl yangkit-data-json-codec -am -Dtest=AnydataValidationOptionsJsonCodecTest,JsonCodecNotificationsTest -Dsurefire.failIfNoSpecifiedTests=false test

[lllyfeng]

#50 is now merged. @Deepya1502, could you please rebase this onto the latest main and resolve the conflicts? Also, a reminder of the open question from my earlier review: why does getSchemaPath() return null for this list? It should already be set by the SCHEMA_TREE phase, so I'd like to understand the root cause before approving. Recording a validation error instead of throwing IllegalArgumentException when no ancestor is found would also match the rest of the codebase better.

[Deepya1502]

Thanks for the comment @lllyfeng , You’re right: this list should already have a schema path after  SCHEMA_TREE . I traced the null back to the inheritance-to-composition refactor in  c18dd66d :  DataDefinitionImpl  delegates schema getters to  SchemaNodeSupport , but its build lifecycle never calls  SchemaNodeSupport.buildSelf() . The phase completes without assigning paths. The paths are already null before anydata deserialization, and the codec reuses the original schema nodes, so this is not specific to anydata.
I’ve prepared a local fix restoring that lifecycle delegation and removed the ancestor-walking workaround. I also addressed the related lifecycle gaps for operations and structures. Unique validation now records an  OPERATION_FAILED  error for missing paths or invalid ancestry rather than throwing, while continuing validation.
Regression tests cover the expected schema paths, duplicate/distinct unique values, missing paths, invalid ancestry, and continued validation. These and the full Maven package build pass locally.
Rebased the code onto latest main to resolve conflicts.

[lllyfeng]

Thanks @Deepya1502, great root-cause analysis. Restoring the SchemaNodeSupport build lifecycle fixes the null schema paths introduced by c18dd66, which also caused an NPE in unique validation for any list on main. The tests fail on main and pass here, and CI is green. Approving and merging.

Two non-blocking follow-ups:

1. Please update the PR title/description: the ancestor-walking workaround is gone, so it should describe restoring the schema path build lifecycle and fixing the unique-validation NPE (regression from c18dd66).
2. When a unique constraint can't be resolved, matchUnique records the same OPERATION_FAILED once per list entry. It would be cleaner to resolve the unique paths once before the entry loop and record a single error. A small follow-up PR would be welcome.