跳到主要内容

安装与配置

引入依赖​

在使用 jdbc-milvus 之前，需要在项目中引入相应的 Maven 依赖。当前版本：6.7.0

<dependency>
    <groupId>net.hasor</groupId>
    <artifactId>jdbc-milvus</artifactId>
    <version>6.7.0</version> <!-- 请使用最新版本 -->
</dependency>

JDBC 连接​

使用标准的 JDBC URL 格式进行连接。

jdbc:dbvisitor:milvus://<host>:<port>/<database>?<parameters>
  • host：Milvus 服务器地址。
  • port：Milvus gRPC 端口（默认 19530）。
  • database：数据库名称（默认 default）。
  • parameters：连接参数，详见 连接参数。

集群模式下多个主机用 ; 分隔，可附带健康检测端口：

jdbc:dbvisitor:milvus://host1:19530:9091;host2:19530:9091/default

示例代码​

1. 创建集合与索引​

String url = "jdbc:dbvisitor:milvus://127.0.0.1:19530/default";
try (Connection conn = DriverManager.getConnection(url)) {
    try (Statement stmt = conn.createStatement()) {
        // 创建集合（Collection）
        stmt.executeUpdate("CREATE TABLE book_vectors ("
                + "book_id int64 PRIMARY KEY, "
                + "word_count int64, "
                + "book_intro float_vector(2)"
                + ")");

        // 创建向量索引
        stmt.executeUpdate("CREATE INDEX idx_book_intro ON book_vectors (book_intro) "
                + "USING \"IVF_FLAT\" WITH (nlist = 1024, metric_type = \"L2\")");

        // 加载集合到内存（搜索前必须执行）
        stmt.executeUpdate("LOAD TABLE book_vectors");
    }
}

2. 使用 PreparedStatement 插入数据​

String insertSql = "INSERT INTO book_vectors (book_id, word_count, book_intro) VALUES (?, ?, ?)";
try (PreparedStatement ps = conn.prepareStatement(insertSql)) {
    ps.setLong(1, 1L);
    ps.setLong(2, 1000L);
    ps.setObject(3, Arrays.asList(0.1f, 0.2f)); // 向量使用 List<Float>
    ps.executeUpdate();
}

3. 向量搜索（KNN）​

使用 ORDER BY vector_field <-> vector LIMIT n 语法执行最近邻搜索。

String searchSql = "SELECT book_id, word_count FROM book_vectors "
        + "ORDER BY book_intro <-> ? LIMIT 5";
try (PreparedStatement ps = conn.prepareStatement(searchSql)) {
    ps.setObject(1, Arrays.asList(0.1f, 0.2f));
    try (ResultSet rs = ps.executeQuery()) {
        while (rs.next()) {
            System.out.println("ID: " + rs.getLong("book_id")
                    + ", Count: " + rs.getLong("word_count"));
        }
    }
}

4. 向量范围搜索​

使用 WHERE vector_field <-> vector < radius 或 vector_range 函数执行范围搜索。

// 方式 1：比较表达式
String rangeSql1 = "SELECT * FROM book_vectors WHERE book_intro <-> [0.1, 0.2] < 0.5";

// 方式 2：vector_range 函数（推荐）
String rangeSql2 = "SELECT * FROM book_vectors WHERE vector_range(book_intro, [0.1, 0.2], 0.5)";

try (Statement stmt = conn.createStatement()) {
    try (ResultSet rs = stmt.executeQuery(rangeSql2)) {
        while (rs.next()) {
            System.out.println("ID: " + rs.getLong("book_id"));
        }
    }
}

5. 标量过滤查询​

try (Statement stmt = conn.createStatement()) {
    try (ResultSet rs = stmt.executeQuery(
            "SELECT book_id, word_count FROM book_vectors WHERE word_count > 500 LIMIT 10")) {
        while (rs.next()) {
            System.out.println("ID: " + rs.getLong("book_id"));
        }
    }
}

6. 混合查询（标量过滤 + 向量搜索）​

String sql = "SELECT book_id FROM book_vectors "
        + "WHERE word_count > 500 "
        + "ORDER BY book_intro <-> [0.1, 0.2] LIMIT 3";
try (Statement stmt = conn.createStatement()) {
    try (ResultSet rs = stmt.executeQuery(sql)) {
        while (rs.next()) {
            System.out.println("ID: " + rs.getLong("book_id"));
        }
    }
}

7. 使用 Hint​

通过 SQL Hint 覆盖查询行为。

try (Statement stmt = conn.createStatement()) {
    // 覆盖 LIMIT 和 OFFSET
    ResultSet rs = stmt.executeQuery(
            "/*+ overwrite_find_limit=5, overwrite_find_skip=10 */ "
            + "SELECT * FROM book_vectors WHERE word_count > 0");

    // 将查询转换为 COUNT
    ResultSet countRs = stmt.executeQuery(
            "/*+ overwrite_find_as_count=true */ "
            + "SELECT * FROM book_vectors WHERE word_count > 500");
}