Fix ArrayIndexOutOfBoundsException in CompositeBitmapIndex null handling

[hrstoyanov]

Summary

Fixes ArrayIndexOutOfBoundsException in AbstractCompositeBitmapIndex.internalHandleChanged() when updating an entity from null to non-null composite key value (or vice versa).

Issue Description

When an entity with a composite-indexed field (e.g., IndexerInstant for Instant fields) is:

1. Added with a null value → composite key = NULL() = Object[1]
2. Updated to a non-null value → composite key = Object[6] (for Instant's 6 components)

The method would throw:

java.lang.ArrayIndexOutOfBoundsException: Index 1 out of bounds for length 1
    at org.eclipse.store.gigamap.types.SubBitmapIndexHashing.index()

Root Cause

The original internalHandleChanged() method called ensureSubIndices(newKeys) which grew the subIndices array, then iterated all sub-indices calling internalHandleChanged(oldKeys, ...). When oldKeys was shorter than the new array (e.g., length 1 vs 6), accessing oldKeys[position] threw ArrayIndexOutOfBoundsException.

Solution

Modified internalHandleChanged() to check isEmpty(oldKeys, i) and isEmpty(newKeys, i) per sub-index position:

• Both empty: Skip (nothing to do)
• Old empty, new non-empty: ADD via internalAddToEntry()
• Old non-empty, new empty: REMOVE via internalRemove()
• Both non-empty: HANDLE CHANGE via internalHandleChanged()

Files Changed

1. AbstractCompositeBitmapIndex.java - Fixed internalHandleChanged() method (lines 529-568)
2. CompositeIndexNullHandlingTest.java (NEW) - Comprehensive test coverage with 5 test methods

Testing

Test covers:

• null→non-null updates
• non-null→null updates
• Persistence with EmbeddedStorage
• Multiple entity updates
• Query operations with mixed null/non-null values

Impact

• Breaking Changes: None
• Backward Compatible: Yes
• Performance Impact: Minimal (adds 2 boolean checks per iteration)

Discovered by TourBiz project during Booking entity updates with nullable Instant fields.

[hrstoyanov]

This addresses #623

[hrstoyanov]

I have 100+ tests for Eclipse Store and with this fix it all pass now.

Please release 4.0.2

[zdenek-jonas]

Thanks for the PR and the contribution. Due to our current review capacity, this will likely not be reviewed for at least a couple of weeks. Thank you for your patience.

[Copilot]

Pull request overview

Fixes a composite-bitmap-index update bug where transitioning an indexed composite key between null and non-null could trigger ArrayIndexOutOfBoundsException due to mismatched key-array lengths across sub-indices.

Changes:

• Update AbstractCompositeBitmapIndex.internalHandleChanged() to dispatch per sub-index position (add/remove/change/skip) based on old/new key presence.
• Add a new JUnit test suite covering null↔non-null composite key updates, persistence restart, and query behavior.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.

File	Description
gigamap/gigamap/src/main/java/org/eclipse/store/gigamap/types/AbstractCompositeBitmapIndex.java	Fixes update logic to handle old/new composite key arrays of different lengths safely.
gigamap/gigamap/src/test/java/org/eclipse/store/gigamap/indexer/CompositeIndexNullHandlingTest.java	Adds regression tests for null-handling transitions, persistence, multi-entity updates, and queries.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The final assertion expects 2 results for null OR 2025-06-15T10:30:00Z after event1 is updated to 11:30. At that point neither entity has a null timestamp, and only event2 matches 10:30, so the expected count should be 1 (or adjust the predicate to match both entities).

Suggested change

	assertEquals(2, map.query(indexer.is((Instant) null).or(indexer.is(Instant.parse("2025-06-15T10:30:00Z")))).count());
	assertEquals(1, map.query(indexer.is((Instant) null).or(indexer.is(Instant.parse("2025-06-15T10:30:00Z")))).count());

[Copilot]

Unused import: org.eclipse.store.gigamap.types.BitmapIndices is not referenced in this test class. Please remove it to avoid unused-import warnings/failures in stricter builds.

[fh-ms]

Thank you for contributing!

The code looks fine, but the provided test fails: https://github.com/eclipse-store/store/actions/runs/23066299067/job/69194135361?pr=624#step:4:11679

This needs to be addressed before merging, as it is mandatory for you to sign the ECA in order to contribute:
https://www.eclipse.org/projects/handbook/#resources-eca

Please have a look at Copilot's mentions too.

[hrstoyanov]

@fh-ms Tried to create Eclipse account for contributor, but fit fails and does not work ... I fixed the junit test

[fh-ms]

If you have trouble creating an Eclipse account, I'll close this PR and copy your changes to a new one if you are fine with it.

[hrstoyanov]

@fh-ms - Yes, do it, please - anyway we fix this is a good way!