/batch endpoint returns HTTP 500 NullPointerException on commit — but data is committed

[babasaikiran]

Environment

ArcadeDB version: 26.5.1-SNAPSHOT (build 25ed565)
Cluster: 3-node Raft HA

Summary

The /api/v1/batch/{db} endpoint returns HTTP 500 with a NullPointerException after
successfully committing all vertices and edges. The response body reports:

{
  "error": "Error on transaction commit",
  "detail": "Cannot invoke \"com.arcadedb.database.Binary.toByteArray()\" because \"phase1.result\" is null",
  "exception": "java.lang.NullPointerException"
}

However, SELECT count(*) confirms that all records were persisted correctly.

Steps to Reproduce

python reproduce_batch_npe.py

#!/usr/bin/env python3

import json
import requests
from requests.auth import HTTPBasicAuth

ARCADE_URL = "http://10.91.15.22:2480/api/v1"
DB = "batch_npe_repro"
USER = "root"
PASSWORD = "pass"
NUM_VERTICES = 500
NUM_EDGES = 500

session = requests.Session()
session.auth = HTTPBasicAuth(USER, PASSWORD)


def sql(statement):
    r = session.post(
        f"{ARCADE_URL}/command/{DB}",
        json={"language": "sql", "command": statement},
        timeout=30,
    )
    return r.status_code, r.json() if r.status_code == 200 else r.text


def create_database():
    session.post(f"{ARCADE_URL}/server", json={"command": f"drop database {DB}"}, timeout=30)
    r = session.post(f"{ARCADE_URL}/server", json={"command": f"create database {DB}"}, timeout=30)
    if r.status_code != 200:
        raise RuntimeError(f"Cannot create database: {r.status_code}")


def setup_schema():
    sql("CREATE VERTEX TYPE TestNode IF NOT EXISTS")
    sql("CREATE PROPERTY TestNode.node_id IF NOT EXISTS STRING")
    sql("CREATE INDEX ON TestNode (node_id) UNIQUE")
    sql("CREATE EDGE TYPE TestEdge IF NOT EXISTS")


def build_payload():
    lines = []
    for i in range(NUM_VERTICES):
        lines.append(json.dumps({
            "@type": "vertex", "@class": "TestNode", "@id": f"n{i}",
            "node_id": f"n{i}", "label": f"Node {i}",
        }))
    for i in range(NUM_EDGES):
        lines.append(json.dumps({
            "@type": "edge", "@class": "TestEdge",
            "@from": f"n{i}", "@to": f"n{(i + 1) % NUM_VERTICES}", "weight": i,
        }))
    return "\n".join(lines)


def send_batch(payload):
    resp = session.post(
        f"{ARCADE_URL}/batch/{DB}?batchSize=50000&commitEvery=50000",
        data=payload.encode("utf-8"),
        headers={"Content-Type": "application/x-ndjson"},
        timeout=120,
    )
    return resp.status_code, resp.text


def verify():
    _, v = sql("SELECT count(*) AS c FROM TestNode")
    _, e = sql("SELECT count(*) AS c FROM TestEdge")
    v_count = v.get("result", [{}])[0].get("c", 0) if isinstance(v, dict) else -1
    e_count = e.get("result", [{}])[0].get("c", 0) if isinstance(e, dict) else -1
    return v_count, e_count


def main():
    create_database()
    setup_schema()
    status, body = send_batch(build_payload())
    v_count, e_count = verify()

    if status == 500 and v_count == NUM_VERTICES and e_count == NUM_EDGES:
        print(f"HTTP 500 but all data committed. vertices={v_count}/{NUM_VERTICES}, edges={e_count}/{NUM_EDGES}, response={body[:300]}")
    elif status != 200:
        print(f"HTTP {status} — vertices={v_count}, edges={e_count}, response={body[:300]}")
    elif v_count != NUM_VERTICES or e_count != NUM_EDGES:
        print(f"Count mismatch: vertices={v_count}/{NUM_VERTICES}, edges={e_count}/{NUM_EDGES}")


if __name__ == "__main__":
    main()

Expected Behavior

The /batch endpoint should return HTTP 200 with accurate verticesCreated / edgesCreated
counts when the data is successfully committed.

Actual Behavior

HTTP 500 is returned with a NullPointerException on phase1.result. The data is committed
to disk, but the client receives an error response.

[lvca]

Already fixed today by 7fb46fc. Added a regression test.