ストアドプロシージャとUDF

Welcome to extending Neo4j! Sometimes, Cypher isn't enough for complex logic or specific integrations. That's where custom code comes in.

Neo4j allows you to extend its functionality using Stored Procedures and User-Defined Functions (UDFs), typically written in Java.

• Stored Procedures perform actions, like creating nodes or running complex algorithms.
• User-Defined Functions (UDFs) return values, just like built-in Cypher functions.

They encapsulate complex logic, improve query readability, and boost performance for repetitive tasks.

It's crucial to understand the difference between procedures and UDFs:

• Stored Procedures:
  - Called with CALL package.procedure()
  - Can perform side effects (create, update, delete data)
  - Return tabular results (rows and columns)
  - Can access the graph database directly
• User-Defined Functions (UDFs):
  - Called within Cypher expressions (e.g., RETURN my.udf(n.property))
  - Must be pure functions (no side effects)
  - Return a single scalar value (e.g., string, number, boolean, list)
  - Cannot modify the graph or access it directly

Stored procedures are Java classes compiled into a JAR file. They use specific Neo4j annotations.

Here's the basic structure for a simple procedure that doesn't take any parameters and returns a string:

Let's break down the previous code:

• @Procedure(value = "my.hello"): This annotation declares a method as a stored procedure and defines its full name (my.hello).
• @Description: Provides a description visible in Neo4j Browser.
• Output Class (StringOutput): Procedures return a Stream of custom objects. Each object represents a row in the result, and its public fields become the column names.
• Stream<StringOutput>: The return type for procedures.

This Java code is ready to be compiled into a JAR and deployed.

To make your Java procedure available in Neo4j, you need to compile it and place the resulting JAR file into the database's plugins directory.

Steps:

1. Compile your Java code into a JAR file (e.g., my-procedures.jar).
2. Copy the JAR file into your Neo4j installation's plugins folder.
3. Restart your Neo4j database instance.

After restarting, Neo4j will discover and register your new procedures and functions.

Once deployed, you can call your procedure using the CALL keyword in Cypher. Let's try calling our my.hello procedure:

UDFs are similar to procedures but have different annotations and return types. They are designed to be used inline within Cypher expressions.

Here's the basic structure for a UDF that takes a string and returns a modified string:

Let's look at the UDF's components:

• @UserFunction("my.capitalize"): This annotation declares a method as a UDF and defines its full name (my.capitalize).
• @Name("input"): Specifies the name for the parameter when used in Cypher.
• Return Type: UDFs return a single scalar value (e.g., String, Long, Boolean, List<String>). They do not return a Stream or custom output objects.

Like procedures, this Java code must be compiled into a JAR and deployed to the plugins folder.

After deploying your UDF, you can use it directly within Cypher queries as part of an expression. It behaves just like built-in functions.

Let's use our my.capitalize UDF:

You've learned how to extend Neo4j with custom Java code!

• Stored Procedures execute complex actions, return tabular results, and are called with CALL.
• User-Defined Functions (UDFs) return single scalar values, are pure functions (no side effects), and are used inline in Cypher expressions.
• Both require Java code, specific Neo4j annotations, compilation into a JAR, and deployment to the Neo4j plugins directory.

These powerful extensions allow you to integrate custom logic, algorithms, and external services directly into your Neo4j environment, significantly expanding its capabilities.