[scottf]

Two read-side fixes in NatsObjectStore.get.

Empty object. put of an empty stream stores an object with 0 chunks. get only special-cased 1 chunk, so for 0 chunks it subscribed and waited in nextMessage(jsm.getTimeout()) for chunks that do not exist, then passed the size and digest checks. Every get of an empty object took the full JetStream timeout (2 s by default). The chunk read is now skipped when the object has no chunks.

Single chunk with no data. Without direct get, the message get API returns no data for an empty payload, so MessageInfo.getData() is null. The single-chunk branch passed it to digester.update(...), which threw a NullPointerException. The data is now digested and written only when present, so the existing size check reports the mismatch.

Tests, in ObjectStoreTests:

• testEmptyObject: puts an empty object, asserts get returns before the JetStream timeout with size 0 and 0 chunks. Without the fix it fails, each get taking 2.1 to 2.3 s.
• testSingleChunkWithNoData: puts a one-chunk object, turns off allowDirect on the stream, publishes an empty payload as the chunk, and asserts get throws the OsGetSizeMismatch error. Without the fix it fails with NullPointerException.

ObjectStoreTests: 13 tests, 0 failures.

🤖 Generated with Claude Code

[synadia-claude-reviewer]

Claude finished @scottf's task in 58s —— View job

---

Review complete

• Fetch PR diff and context
• Review NatsObjectStore.get changes for correctness
• Review new tests
• Post review summary

Summary: Both fixes in NatsObjectStore.get (src/main/java/io/nats/client/impl/NatsObjectStore.java) are correct and well-scoped.

• Empty object skip: Changing the fallback branch to else if (oi.getChunks() > 0) correctly avoids subscribing/waiting when there are no chunks. The post-loop checks (totalChunks != oi.getChunks(), totalBytes != oi.getSize(), digest match) still hold with totalChunks/totalBytes left at 0, since put always computes a digest (even over zero bytes) via digester.getDigestEntry() — so testEmptyObject's digest/size assertions are consistent with put's behavior.
• Null-data NPE fix: MessageInfo.getData() is annotated @Nullable (confirmed in MessageInfo.java), so guarding digester.update(data)/out.write(data) with data != null is the right fix. Note the multi-chunk subscription loop (lines ~239-243) doesn't need the same guard — it reads from a regular Message, not MessageInfo, and a normal NATS protocol payload is never null (zero-length array at worst), so there's no parallel NPE risk there.
• Size mismatch still correctly surfaces as OsGetSizeMismatch when a chunk's actual data length doesn't match the recorded object size, which is what testSingleChunkWithNoData verifies.

No security, thread-safety, or API-compatibility concerns. Tests look solid and exercise both fixed paths directly.