[FlatPostgresCollection] Sort keys before preparedStatement.addBatch() to prevent deadlocks

[suddendust]

Sort keys to prevent deadlocks in batched writes on FlatPostgresCollection

Summary

Enforces a canonical, string-sorted key order across all JDBC-batched, key-keyed write operations in FlatPostgresCollection. Postgres acquires per-row locks in batch-entry order for INSERT ... ON CONFLICT DO UPDATE and UPDATE ... WHERE key = ?, so two concurrent batches that touch overlapping key sets in opposite orders can form a wait-for cycle and one side gets aborted with ERROR: deadlock detected (SQLState 40P01).

Sorting inside the store — the last hop before addBatch(...) — makes the invariant local to the collection and independent of caller iteration order, so no downstream service has to remember to sort.

Symptom this fixes

Callers were seeing production log lines like:

Failed to execute batch update. SQL: UPDATE "myTable" SET "attributes.col1" = ?, "col2" = ? WHERE "pkCol" = ?
Error: Batch entry 8 UPDATE "myTable" ... was aborted:
ERROR: deadlock detected.

The offending SQL is produced only by the Map-based bulkUpdate on this class, and the deadlock reproduces trivially with two concurrent callers that iterate overlapping keys in opposite orders. The same class of bug also exists on bulkUpsert / bulkCreateOrReplace, which batch INSERT ... ON CONFLICT DO UPDATE — the DO UPDATE branch takes a row-exclusive lock and follows the same batch-entry ordering.

Testing

Didn't add any additional tests - The existing integration tests should be sufficient.

Checklist:

• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• Any dependent changes have been merged and published in downstream modules

[suddendust]

This array contains the order of the keys - We add to the batch according to this order.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 80.90%. Comparing base (b381c1b) to head (4b37f48).

Additional details and impacted files

@@             Coverage Diff              @@
##               main     #320      +/-   ##
============================================
+ Coverage     80.89%   80.90%   +0.01%     
  Complexity     1580     1580              
============================================
  Files           242      242              
  Lines          7628     7632       +4     
  Branches        751      752       +1     
============================================
+ Hits           6171     6175       +4     
  Misses          968      968              
  Partials        489      489              

Flag	Coverage Δ
integration	80.90% <100.00%> (+0.01%)	⬆️
unit	55.31% <14.28%> (-0.03%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

Test Results

  124 files  ±0    124 suites  ±0   38s ⏱️ ±0s
  846 tests ±0    845 ✅ ±0  1 💤 ±0  0 ❌ ±0 
1 183 runs  ±0  1 182 ✅ ±0  1 💤 ±0  0 ❌ ±0 

Results for commit 4b37f48. ± Comparison against base commit b381c1b.

♻️ This comment has been updated with latest results.