Lynxus is a pragmatic MyBatis alternative. Your SQL is validated at compile time, your mappers become ordinary Java classes, and runtime is one explicit JDBC contract.
Lynxusio.github.lynxus:lynxus-spring-boot-starterpackage com.example.user;import io.github.lynxus.annotation.*;@Mapperpublic interface UserMapper { @Select("SELECT id, name FROM users WHERE id = #{userIdid}") User findById(@Param("id") Long id); @Insert(""" INSERT INTO users (id, name) VALUES (#{id}, #{name})""") int insert(@Param("id") Long id, @Param("name") String name);}
final class UserMapperImpl implements UserMapper { private final SqlExecutor executor; UserMapperImpl(SqlExecutor executor) { this.executor = executor; } @Override public User findById(Long id) { return executor.query(FIND_BY_ID.bind(id)).oneOrNull(); } private static final QueryDefinition<User> FIND_BY_ID = QueryDefinition.assembled( "UserMapper#findById", SQL, ROW_MAPPER);}
A mistyped parameter is a build error, not a production incident.
01 How it works
The mapper stays the source of truth. Everything else — validation, implementations, the mapper index — is produced by javac.
package com.example.user;@Mapperpublic interface UserMapper { @Select("SELECT id, name FROM users WHERE id = #{userIdid}") User findById(@Param("id") Long id); @Insert("INSERT INTO users (id, name) VALUES (#{id}, #{name})") int insert(@Param("id") Long id, @Param("name") String name);}
$ mvn -q compile[lynxus] 1 mapper interface · 2 statements[lynxus] UserMapper#findById SQL parsed · 1 bound parameter: id result mapping: User ← id, name[lynxus] wrote UserMapperImpl.java[lynxus] wrote META-INF/lynxus/mapper-index.jsonBUILD SUCCESS 0.41s
final class UserMapperImpl implements UserMapper { private final SqlExecutor executor; @Override public User findById(Long id) { return executor.query(FIND_BY_ID.bind(id)).oneOrNull(); } private static final QueryDefinition<User> FIND_BY_ID = QueryDefinition.assembled( "UserMapper#findById", SQL, ROW_MAPPER);}
var executor = JdbcAssembly.create(dataSource);var mapper = new UserMapperImpl(executor);User user = mapper.findById(42L);// → PreparedStatement.executeQuery()// no proxy · no XML · no reflection
SQL sits next to the method that uses it — no XML file to hunt down, no namespace to keep in sync.
02 Side by side
The migration is mostly deletion: the XML goes away, the runtime parser goes away, and the mapper you already wrote stays.
<!-- resources/mapper/UserMapper.xml --><mapper namespace="com.example.UserMapper"> <select id="findById" resultType="User"> SELECT id, name FROM users <where> <if test="id != null">AND id = #{id}</if> </where> </select></mapper>
@Mapperpublic interface UserMapper { @Select("SELECT id, name FROM users WHERE id = #{id}") User findById(@Param("id") Long id);}
Same SQL. Same result. One of them fails before your app starts.
03 What is not there
No parser to boot, no expression language to debug, no proxies to explain in a stack trace. Startup reads a generated index; every call is a prepared statement and a row mapper you can open in your IDE.
AOT & Native Image guide04 Get started
Java 21, one dependency, and a mapper you already know how to write.
io.github.lynxus:lynxus-spring-boot-starter