Neo4J自定义存储过程调用含Periodic Commit的查询报错

Hey there! Let's break down why your custom stored procedure is throwing errors when called via db.execute() in Neo4j Desktop, even though it works in unit tests and the query runs fine on its own. Here are the most likely culprits and actionable fixes:

1. Permission & File Access Mismatches #

This is the top suspect. When you run the query manually in Neo4j Desktop, you're operating under your user account's permissions—but custom procedures run under the Neo4j process's context, which has stricter restrictions:

• CSV File Location: Move your CSV to Neo4j's designated import directory (default path: neo4j-home/import). Reference it in your query using the file:/// protocol (e.g., file:///my_nodes.csv instead of a local filesystem path like /Users/you/data.csv).
• Procedure Unrestricted Access: Open your neo4j.conf file and add your procedure's package to the unrestricted list. For example:

  dbms.security.procedures.unrestricted=com.yourcompany.procedures.*
  

  Restart Neo4j after making this change.

2. Execution Context Differences #

The db.execute() call inside your procedure runs in a distinct context compared to manual query execution:

• Transaction Boundaries: If your procedure is annotated with @Procedure(mode = Mode.WRITE), it’s already running in a write transaction. Calling db.execute() might trigger unexpected conflicts. Try using the procedure’s existing transaction context instead, or explicitly manage transactions if needed.
• Path Resolution: Relative paths that work in manual queries might not resolve correctly when run via the procedure. Always use absolute paths for CSV files, or validate paths against Neo4j’s allowed directories first.

3. Missing Error Details & Logging #

Unit tests often show full error traces, but Neo4j Desktop might simplify exceptions from procedures. Add explicit logging and error handling to uncover the root cause:

import org.neo4j.logging.Log;
import java.sql.SQLException;
import java.util.Map;

public class YourImportProcedure {
    private final Log log;

    public YourImportProcedure(Log log) {
        this.log = log;
    }

    @Procedure(name = "com.yourcompany.importFromExternal", mode = Mode.WRITE)
    public void importNodes() throws SQLException {
        String csvPath = "file:///imported_nodes.csv";
        String query = "LOAD CSV FROM $path AS row CREATE (:ExternalNode {id: row[0], data: row[1]})";
        
        log.info("Starting import with query: {}", query);
        log.info("Using CSV path: {}", csvPath);

        try (var stmt = db.execute(query, Map.of("path", csvPath))) {
            log.info("Import completed successfully, created {} nodes", stmt.getUpdateCount());
        } catch (SQLException e) {
            log.error("Import failed with error: {}", e.getMessage(), e);
            throw new RuntimeException("Failed to import nodes", e);
        }
    }
}

Check Neo4j’s logs (in Desktop, go to your database → "Logs" tab) for the full error stack trace—it will point you exactly to what’s breaking.

4. Environment Discrepancies Between Tests & Desktop #

Your unit tests likely use an embedded Neo4j instance, which has different defaults than the standalone Desktop setup:

• File System Access: Embedded instances might have full local filesystem access, while the Desktop process could be sandboxed (especially on macOS/Windows).
• Configuration: Unit tests might override settings like dbms.security.allow_csv_import_from_file_urls that aren’t enabled in your Desktop config. Double-check these in neo4j.conf.

Quick Fix Checklist #

1. Move your CSV to Neo4j’s import directory and update the query path to use file:///.
2. Add your procedure’s package to dbms.security.procedures.unrestricted in neo4j.conf and restart.
3. Add logging to your procedure to capture the exact error and query being executed.
4. Verify the Neo4j process has read permissions on the CSV file and its parent directory.

内容的提问来源于stack exchange，提问作者matentzn