导入Neo4j的方式有三种，大概如下：

1. 读取外部文件到内存中，然后使用create 语句导入之后建立关系。

2. 用load csv 读取csv 文件

3. 从JDBC直接load 到neo4j 中

这里测试导入的场景如下：

主要包含四个字段： 卡号|用户名称|转账卡号|转让金额；主要逻辑是： 采用序号递增，从0 - 10000，用户名称也是从"user" + 0-10000。转账卡号是转给下一个节点，转账金额也是递增。

　　下面的构造数据统一采用从内存中构造数据，构造1W条数据，也就是1W个node，1W条关系。

    private static final Integer DATA_SIZE = 10000;

    private static List<Map<String, Object>> generateData() {
        List<Map<String, Object>> datas = new ArrayList<>(DATA_SIZE);
        Map<String, Object> tmpMap = null;
        for (int i = 0; i < DATA_SIZE; i++) {
            tmpMap = new HashMap<>();
            datas.add(tmpMap);
            tmpMap.put("cardNum", i);
            tmpMap.put("userName", "user" + i);
            tmpMap.put("transferCardNum", (i + 1) % 10000); // 每个卡给自己的下一个卡转钱
            tmpMap.put("transferAmount", i);
        }
        return datas;
    }

0. pom 引入如下依赖

        <!-- neo4j 相关的API -->
        <dependency>
            <groupId>org.neo4j.driver</groupId>
            <artifactId>neo4j-java-driver</artifactId>
            <version>4.0.0</version>
        </dependency>
        <dependency>
            <groupId>org.neo4j</groupId>
            <artifactId>neo4j</artifactId>
            <version>3.3.4</version>
        </dependency>

1. 外部文件导入neo4j

　　这里导入文件读入文件的过程忽略掉，从内存中模拟1W条数据。

    /**
     * 测试手动插入数据以及维护关系
     */
    private static void inertNeo4jTest() {
        // 构造数据， 数据和pg 库里面的数据一样
        List<Map<String, Object>> datas = generateData();

        Driver driver = GraphDatabase.driver("bolt://localhost:7687", AuthTokens.basic("neo4j", "neo4j."));
        Session session = driver.session();

        StopWatch stopWatch = new StopWatch();
        stopWatch.start();
        // 手动create to neo4j
        String createCQLTemplate = "create (n:transferDetail {cardNum: '$cardNum', userName: '$userName', transferCardNum: '$transferCardNum', transferAmount: '$transferAmount'})";
        datas.forEach(data -> {
            String createCQL = createCQLTemplate.replace("$cardNum", MapUtils.getString(data, "cardNum"))
                    .replace("$userName", MapUtils.getString(data, "userName"))
                    .replace("$transferCardNum", MapUtils.getString(data, "transferCardNum"))
                    .replace("$transferAmount", MapUtils.getString(data, "transferAmount"));
            session.run(createCQL);
        });
        System.out.println("插入成功耗时： " + stopWatch.getTime() + " ms");
        // 手动维护关系
        String mergeCQLTemplate = "match (a:transferDetail{cardNum: '$cardNum1'}), (b:transferDetail{cardNum: '$cardNum2'}) MERGE(a)-[:TRANSFER{transferAmount: '$transferAmount'}]->(b)";
        datas.forEach(data -> {
            String mergeCQL = mergeCQLTemplate.replace("$cardNum1", MapUtils.getString(data, "cardNum"))
                    .replace("$cardNum2", MapUtils.getString(data, "transferCardNum"))
                    .replace("$transferAmount", MapUtils.getString(data, "transferAmount"));
            session.run(mergeCQL);
        });
        stopWatch.stop();
        System.out.println("转换关系成功，耗时： " + stopWatch.getTime() + " ms");

        // close resource
        session.close();
        driver.close();
    }

　　代码逻辑很简单，从内存构造1W条数据-》create 到 neo4j -》手动维护关系。 测试结果耗时如下：

插入成功耗时： 50976 ms
转换关系成功，耗时： 285111 ms

可以看到消耗时间大概为不到5min。到neo4j 查看数据如下：

MATCH (n:transferDetail) RETURN count(n)

2. 测试load csv

1. pom 引入如下依赖：

        <dependency>
            <groupId>org.apache.commons</groupId>
            <artifactId>commons-csv</artifactId>
            <version>1.3</version>
        </dependency>

2. 测试方法： 生成csv 文件，然后load csv

    private static void loadCSVtest() throws Exception {
        // 读取CSV 文件
//        Reader fileReader = new FileReader(fileName);
//        Iterable<CSVRecord> records = CSVFormat.RFC4180.withFirstRecordAsHeader().parse(fileReader);
//        for (CSVRecord record : records) {
//            System.out.println(record.get("instanceId") + record.get("regionId") + record.get("zoneId"))
//        }
        // 1. 写入一个csv 到本地， 构造相同的数据
        StopWatch stopWatch = new StopWatch();
        stopWatch.start();
        Appendable fileWriter = new FileWriter("E:\\neo4j3.5\\neo4j-community-3.5.5\\import\\transfer.csv");
        CSVPrinter printer = CSVFormat.RFC4180.withHeader("cardnum", "username", "transfercardnum", "transferamount").print(fileWriter);
        List<Map<String, Object>> datas = generateData();
        datas.forEach(data -> {
            try {
                printer.printRecord(MapUtils.getString(data, "cardNum"), MapUtils.getString(data, "userName"), MapUtils.getString(data, "transferCardNum"), MapUtils.getString(data, "transferAmount"));
            } catch (IOException ignore) {
                // ignore
            }
        });
        printer.close();
        System.out.println("csv 文件输出完成， 耗时： " + stopWatch.getTime());

        // 2. csv load 到neo4j
        Driver driver = GraphDatabase.driver("bolt://localhost:7687", AuthTokens.basic("neo4j", "neo4j."));
        Session session = driver.session();
        String deleteCQL = "match (n:transferdetail) detach delete n";
        session.run(deleteCQL);
        System.out.println("neo4j 清空数据库成功， 耗时: " + stopWatch.getTime() + "ms");
        String constraintCQL = "create constraint on (n:transferdetail) ASSERT n.cardnum is unique";
        session.run(constraintCQL);
        String createCQL = "load csv WITH HEADERS from 'file:///transfer.csv' as line create(n:transferdetail{cardnum:line.cardnum, username:line.username, transfercardnum:line.transfercardnum, transferamount:line.transferamount})";
        session.run(createCQL);
        String relateCQL = "load csv WITH HEADERS from 'file:///transfer.csv' as row match(n:transferdetail{cardnum:row.cardnum}),(m:transferdetail{cardnum:row.transfercardnum}) merge (n)-[:transfer{transferamount:row.transferamount}]->(m)";
        session.run(relateCQL);
        stopWatch.stop();
        System.out.println("load csv导入成功， 耗时: " + stopWatch.getTime() + "ms");
        session.close();
        driver.close();

    }

结果： 可以看到非常的快，比自己手动创建节点然后建立关系快多了

csv 文件输出完成， 耗时： 269
neo4j 清空数据库成功， 耗时: 2730ms
load csv导入成功， 耗时: 3070ms

3. 测试从RDBMS 从加载数据

　　这里采用apoc 从jdbc 加载数据。这里采用从PG数据库加载数据。

1. 首先下载apoc 插件

https://github.com/neo4j-contrib/neo4j-apoc-procedures/releases

2. 将下载的jar 包和pg 的驱动包放到 %neo4j%\plugins 目录下，如下：

3. 修改%neo4j-community-3.5.5%\conf文件夹下面neo4j.conf文件， 最后增加如下配置：

dbms.security.procedures.unrestricted=apoc.*
apoc.export.file.enabled=true

4. 重启neo4j server

5. 查看apoc 版本： 查看到版本证明apoc 插件安装成功

return apoc.version()

结果：

 6. 修改程序采用apoc 从jdbc load 数据

(1) 增加pg 驱动

        <dependency>
            <groupId>org.postgresql</groupId>
            <artifactId>postgresql</artifactId>
            <version>42.2.2</version>
        </dependency>

(2) 编写测试类

    public static final void inertFromPGTest() throws Exception {
        // 1. 构造数据
        List<Map<String, Object>> datas = generateData();

        // 2. 插入到PG库
        StopWatch stopWatch = new StopWatch();
        stopWatch.start();
        Connection connection = null;
        try {
            Class.forName("org.postgresql.Driver");
            connection = DriverManager
                    .getConnection("jdbc:postgresql://127.0.0.1:5432/qlq_test",
                            "postgres", "postgres");
        } catch (Exception e) {
            e.printStackTrace();
        }
        Assert.notNull(connection, "链接失败");
        java.sql.Statement statement = connection.createStatement();
        // 清空数据库
        String trunacteSQL = "truncate table transferdetail";
        statement.execute(trunacteSQL);
        System.out.println("删除pg数据库成功， 耗时: " + stopWatch.getTime() + "ms");
        // 插入数据
        List<String> insertSQLs = new ArrayList<>();
        String valueSQL = "('$cardNum', '$userName', '$transferCardNum', '$transferAmount')";
        datas.forEach(data -> {
            insertSQLs.add(valueSQL.replace("$cardNum", MapUtils.getString(data, "cardNum"))
                    .replace("$userName", MapUtils.getString(data, "userName"))
                    .replace("$transferCardNum", MapUtils.getString(data, "transferCardNum"))
                    .replace("$transferAmount", MapUtils.getString(data, "transferAmount")));
        });
        String sql = "insert into transferdetail(cardNum, userName, transferCardNum, transferAmount) values";
        sql += StringUtils.join(insertSQLs, ",");
        statement.execute(sql);
        System.out.println("插入pg数据库成功， 耗时: " + stopWatch.getTime() + "ms");

        // 3. PG 库转换到neo4j 库
        Driver driver = GraphDatabase.driver("bolt://localhost:7687", AuthTokens.basic("neo4j", "neo4j."));
        Session session = driver.session();
        String deleteCQL = "match (n:transferdetail) detach delete n";
        session.run(deleteCQL);
        System.out.println("neo4j 清空数据库成功， 耗时: " + stopWatch.getTime() + "ms");
        String constraintCQL = "create constraint on (n:transferdetail) ASSERT n.cardnum is unique";
        session.run(constraintCQL);
//        cypher = "CALL apoc.periodic.iterate(\"CALL apoc.load.jdbc('jdbc:postgresql://10.95.24.33:5432/qlq_test?user=postgres&password=afim36owsx&characterEncoding=utf-8',\\\"SELECT * FROM transferdetail\\\")\",\"MERGE(n:transferdetail{cardnum:row.cardnum, username:row.username, transfercardnum:row.transfercardnum, transferamount:row.transferamount}) with * MERGE(m:transferdetail{cardnum:row.transfercardnum, username:row.username, transfercardnum:row.transfercardnum, transferamount:row.transferamount}) with * create p=(n)-[r:transfer{transferamount:row.transferamount}]->(m)\",{batchSize:10000,iterateList:true})";  //连接postgresSQL数据库和设计创建neo4j图数据库数据模型
        String createCQL = "CALL apoc.periodic.iterate(\"CALL apoc.load.jdbc('jdbc:postgresql://10.95.24.33:5432/qlq_test?user=postgres&password=afim36owsx&characterEncoding=utf-8',\\\"SELECT * FROM transferdetail\\\")\",\"create(n:transferdetail{cardnum:row.cardnum, username:row.username, transfercardnum:row.transfercardnum, transferamount:row.transferamount})\",{batchSize:10000,iterateList:true})";  //连接postgresSQL数据库和设计创建neo4j图数据库数据模型
        session.run(createCQL);
        String relateCQL = "CALL apoc.periodic.iterate(\"CALL apoc.load.jdbc('jdbc:postgresql://10.95.24.33:5432/qlq_test?user=postgres&password=afim36owsx&characterEncoding=utf-8',\\\"SELECT * FROM transferdetail\\\")\",\"match(n:transferdetail{cardnum:row.cardnum}),(m:transferdetail{cardnum:row.transfercardnum}) merge (n)-[:transfer{transferamount:row.transferamount}]->(m)\",{batchSize:10000,iterateList:true})";  //连接postgresSQL数据库和设计创建neo4j图数据库数据模型
//        String cypher2 = "CALL apoc.periodic.iterate(\"CALL apoc.load.jdbc('jdbc:postgresql://10.95.24.33:5432/qlq_test?user=postgres&password=afim36owsx&characterEncoding=utf-8',\\\"SELECT * FROM transferdetail\\\")\",\"match(n:transferdetail{cardnum:row.cardnum}),(m:transferdetail{cardnum:row.transfercardnum}) create p=(n)-[r:transfer{transferamount:row.transferamount}]->(m)\",{batchSize:10000,iterateList:true})";  //连接postgresSQL数据库和设计创建neo4j图数据库数据模型
        session.run(relateCQL);
        stopWatch.stop();
        System.out.println("apoc导入成功， 耗时: " + stopWatch.getTime() + "ms");

        session.close();
        driver.close();
    }

测试结果如下：

删除pg数据库成功， 耗时: 546ms
插入pg数据库成功， 耗时: 976ms
neo4j 清空数据库成功， 耗时: 2578ms
apoc导入成功， 耗时: 3013ms

 

　　可以得出结论，从jdbc load 数据和load csv 从性能上差不多；两者都快于手动创建和维护关系。