Fix Arrow FFI hash-to-string resolution for fingerprint terms aggregations

[schenksj]

Summary

• Fix regression where Arrow FFI terms aggregations on string fingerprint fields return raw U64 hash values instead of original strings
• Thread hash_resolution_map from Phase 3 touchup through all Arrow FFI conversion paths
• Applies to single-split, multi-split, and multi-split-multi-agg Arrow FFI methods
• Bump version to 0.32.4

Root Cause

perform_search_async_impl_leaf_response_with_aggregations correctly rewrites terms aggs to use _phash_<field> and builds a hash→string resolution map (Phase 3). The JNI object path used this map, but the Arrow FFI path discarded it — key_to_string(Key::U64(v)) just did v.to_string(), producing raw hashes like "18264002321125465436" instead of "alice".

Test plan

• Verify terms aggregation on string fingerprint fields returns original strings via Arrow FFI
• Verify multi-split terms aggregation merges hash resolution maps correctly
• Verify nested bucket flattening (Terms→Terms) resolves hash keys
• Verify non-fingerprint splits are unaffected (resolution map is None)

🤖 Generated with Claude Code

[claude]

Claude finished @schenksj's task —— View job

---

Code Review Complete ✅

Summary: This PR correctly fixes the Arrow FFI hash resolution regression. The implementation properly threads the hash_resolution_map through all Arrow FFI conversion paths and correctly resolves U64 hash values back to original strings.

Review Results:

• Correctness: ✅ No logic errors found
• Security: ✅ No security issues identified
• Build Issues: ✅ No merge conflicts or build-breaking changes
• Error Propagation: ✅ Proper error handling maintained
• API Consistency: ✅ Consistent with existing patterns

Key Changes Reviewed:

1. Enhanced Arrow FFI functions (aggregation_arrow_ffi.rs):

• ✅ aggregation_result_to_record_batch_with_hash_resolution() correctly accepts hash_resolution_map parameter
• ✅ key_to_string_resolved() properly resolves U64 hash keys using the resolution map
• ✅ Terms aggregation now uses key_to_string_resolved() instead of key_to_string()
• ✅ Nested bucket flattening correctly passes hash resolution through recursion

2. JNI Arrow FFI entry points (jni_agg_arrow.rs):

• ✅ All three methods properly extract hash_resolution_map from SearchWithHashContext
• ✅ Single-split aggregation: Line 127 correctly passes ctx.hash_resolution_map.as_ref()
• ✅ Multi-split aggregation: Lines 245-247 correctly merge hash maps from multiple splits
• ✅ Multi-split multi-agg: Lines 543-545 correctly collect and merge hash maps

3. Hash resolution flow verification:

• ✅ perform_search_async_impl_leaf_response_with_aggregations() builds hash resolution map in Phase 3
• ✅ Hash map is correctly bundled in SearchWithHashContext
• ✅ JNI methods extract the hash map and pass it to Arrow FFI conversion
• ✅ Arrow FFI methods use the map to resolve hash values back to strings

No Issues Found:

• Logic: Hash resolution logic is correct throughout the call chain
• Memory Safety: Proper use of Option<&HashMap<u64, String>> references
• Error Handling: Existing error propagation patterns maintained
• Thread Safety: No thread safety concerns in the changes
• Performance: Efficient hash map merging for multi-split scenarios

The fix correctly addresses the root cause where Arrow FFI paths were discarding the hash resolution map that JNI paths were correctly using.

[claude]

Claude finished @schenksj's task —— View job

---

PR Code Review ✅

Reviewing the PR for:

• Correctness bugs (logic errors, off-by-one, null safety)
• Security issues (injection, unsafe operations, credential handling)
• Merge conflict markers or build-breaking issues
• Missing error propagation (especially JNI error swallowing)
• API consistency with existing patterns in the codebase

Review Results: ✅ No Issues Found

Summary: This PR correctly fixes the Arrow FFI hash resolution regression. The implementation properly threads the hash_resolution_map through all Arrow FFI conversion paths and correctly resolves U64 hash values back to original strings.

Key Changes Reviewed:

1. Enhanced Arrow FFI functions (aggregation_arrow_ffi.rs):

• ✅ aggregation_result_to_record_batch_with_hash_resolution() correctly accepts hash_resolution_map parameter
• ✅ key_to_string_resolved() properly resolves U64 hash keys using the resolution map
• ✅ Terms aggregation now uses key_to_string_resolved() instead of key_to_string()
• ✅ Nested bucket flattening correctly passes hash resolution through recursion

2. JNI Arrow FFI entry points (jni_agg_arrow.rs):

• ✅ All three methods properly extract hash_resolution_map from SearchWithHashContext
• ✅ Single-split aggregation: Line 127 correctly passes ctx.hash_resolution_map.as_ref()
• ✅ Multi-split aggregation: Lines 245-247 correctly merge hash maps from multiple splits
• ✅ Multi-split multi-agg: Lines 543-545 correctly collect and merge hash maps

3. Hash resolution flow verification:

• ✅ perform_search_async_impl_leaf_response_with_aggregations() builds hash resolution map in Phase 3
• ✅ Hash map is correctly bundled in SearchWithHashContext
• ✅ JNI methods extract the hash map and pass it to Arrow FFI conversion
• ✅ Arrow FFI methods use the map to resolve hash values back to strings

4. Comprehensive test coverage (AggregationArrowFfiTest.java):

• ✅ New regression tests specifically validate string key resolution vs. raw U64 hashes
• ✅ Tests cover both simple terms aggregations and terms+sub-aggregations
• ✅ Tests use companion splits with string hash optimization enabled to reproduce the exact regression scenario

Specific Findings:

✅ Correctness:

• No logic errors found
• Hash resolution logic is correct throughout the call chain
• Null safety properly handled with Option<&HashMap<u64, String>>

✅ Security:

• No security issues identified
• Proper memory handling with Arrow FFI operations
• Safe use of HashMap references without ownership transfers

✅ Build Issues:

• No merge conflicts or build-breaking changes
• Version bump to 0.32.4 properly applied

✅ Error Propagation:

• Existing error handling patterns maintained throughout
• No JNI error swallowing introduced
• Proper anyhow::Error propagation preserved

✅ API Consistency:

• Changes follow existing patterns in the codebase
• Consistent parameter naming and function signatures
• Proper use of Option<&HashMap<u64, String>> references

The fix correctly addresses the root cause where Arrow FFI paths were discarding the hash resolution map that JNI paths were correctly using. Hash-to-string resolution now works consistently across all aggregation export paths.

---