• 反復執(zhí)行單條插入語句
• xml拼接sql
• 批處理執(zhí)行

先說結論：少量插入請使用反復插入單條數(shù)據(jù)，方便。數(shù)量較多請使用批處理方式。（可以考慮以有需求的插入數(shù)據(jù)量20條左右為界吧，在我的測試和數(shù)據(jù)庫環(huán)境下耗時都是百毫秒級的，方便最重要）無論何時都不用xml拼接sql的方式。

拼接SQL的xml

newId()是sqlserver生成UUID的函數(shù)，與本文內(nèi)容無關

"insertByBatch"?parameterType="java.util.List">
????INSERT?INTO?tb_item?VALUES
????"list"?item="item"?index="index"?separator=",">
????????(newId(),#{item.uniqueCode},#{item.projectId},#{item.name},#{item.type},#{item.packageUnique},
????????#{item.isPackage},#{item.factoryId},#{item.projectName},#{item.spec},#{item.length},#{item.weight},
????????#{item.material},#{item.setupPosition},#{item.areaPosition},#{item.bottomHeight},#{item.topHeight},
????????#{item.serialNumber},#{item.createTime}

Mapper接口Mapper?是?mybatis插件tk.Mapper?的接口，與本文內(nèi)容關系不大

public?interface?ItemMapper?extends?Mapper<Item>?{
????int?insertByBatch(List?itemList);
}

Service類

@Service
public?class?ItemService?{
????@Autowired
????private?ItemMapper?itemMapper;
????@Autowired
????private?SqlSessionFactory?sqlSessionFactory;
????//批處理
????@Transactional
????public?void?add(List?itemList)?{
????????SqlSession?session?=?sqlSessionFactory.openSession(ExecutorType.BATCH,false);
????????ItemMapper?mapper?=?session.getMapper(ItemMapper.class);
????????for?(int?i?=?0;?i?????????????mapper.insertSelective(itemList.get(i));
????????????if(i%1000==999){//每1000條提交一次防止內(nèi)存溢出
????????????????session.commit();
????????????????session.clearCache();
????????????}
????????}
????????session.commit();
????????session.clearCache();
????}
????//拼接sql
????@Transactional
????public?void?add1(List?itemList)?{
????????itemList.insertByBatch(itemMapper::insertSelective);
????}
????//循環(huán)插入
????@Transactional
????public?void?add2(List?itemList)?{
????????itemList.forEach(itemMapper::insertSelective);
????}
}

測試類

@RunWith(SpringRunner.class)
@SpringBootTest(webEnvironment?=?SpringBootTest.WebEnvironment.RANDOM_PORT,?classes?=?ApplicationBoot.class)
public?class?ItemServiceTest?{
????@Autowired
????ItemService?itemService;

????private?List?itemList?=?new?ArrayList<>();
????//生成測試List
????@Before?
????public?void?createList(){
????????String?json?="{\n"?+
????????????????"????????\"areaPosition\":?\"TEST\",\n"?+
????????????????"????????\"bottomHeight\":?5,\n"?+
????????????????"????????\"factoryId\":?\"0\",\n"?+
????????????????"????????\"length\":?233.233,\n"?+
????????????????"????????\"material\":?\"Q345B\",\n"?+
????????????????"????????\"name\":?\"TEST\",\n"?+
????????????????"????????\"package\":?false,\n"?+
????????????????"????????\"packageUnique\":?\"45f8a0ba0bf048839df85f32ebe5bb81\",\n"?+
????????????????"????????\"projectId\":?\"094b5eb5e0384bb1aaa822880a428b6d\",\n"?+
????????????????"????????\"projectName\":?\"項目_TEST1\",\n"?+
????????????????"????????\"serialNumber\":?\"1/2\",\n"?+
????????????????"????????\"setupPosition\":?\"1B柱\",\n"?+
????????????????"????????\"spec\":?\"200X200X200\",\n"?+
????????????????"????????\"topHeight\":?10,\n"?+
????????????????"????????\"type\":?\"Steel\",\n"?+
????????????????"????????\"uniqueCode\":?\"12344312\",\n"?+
????????????????"????????\"weight\":?100\n"?+
????????????????"????}";
????????Item?test1?=?JSON.parseObject(json,Item.class);
????????test1.setCreateTime(new?Date());
????????for?(int?i?=?0;?i?1000;?i++)?{//測試會修改此數(shù)量
????????????itemList.add(test1);
????????}
????}
?????//批處理
????@Test
????@Transactional
????public?void?tesInsert()?{
????????itemService.add(itemList);
????}
????//拼接字符串
????@Test
????@Transactional
????public?void?testInsert1(){
????????itemService.add1(itemList);
????}
????//循環(huán)插入
????@Test
????@Transactional
????public?void?testInsert2(){
????????itemService.add2(itemList);
????}
}

測試結果：

10條 25條數(shù)據(jù)插入經(jīng)多次測試，波動性較大，但基本都在百毫秒級別

其中 拼接sql方式在插入500條和1000條時報錯（似乎是因為sql語句過長，此條跟數(shù)據(jù)庫類型有關，未做其他數(shù)據(jù)庫的測試）：com.microsoft.sqlserver.jdbc.SQLServerException: 傳入的表格格式數(shù)據(jù)流(TDS)遠程過程調(diào)用(RPC)協(xié)議流不正確,此RPC請求中提供了過多的參數(shù),最多應為2100

可以發(fā)現(xiàn)

• 循環(huán)插入的時間復雜度是 O(n),并且常數(shù)C很大
• 拼接SQL插入的時間復雜度（應該）是 O(logn),但是成功完成次數(shù)不多，不確定
• 批處理的效率的時間復雜度是 O(logn)，并且常數(shù)C也比較小

結論

循環(huán)插入單條數(shù)據(jù)雖然效率極低，但是代碼量極少，在使用tk.Mapper的插件情況下，僅需代碼,：

@Transactional
public?void?add1(List?itemList)?{
????itemList.forEach(itemMapper::insertSelective);
}

因此，在需求插入數(shù)據(jù)數(shù)量不多的情況下肯定用它了。

xml拼接sql是最不推薦的方式，使用時有大段的xml和sql語句要寫，很容易出錯，工作效率很低。更關鍵點是，雖然效率尚可，但是真正需要效率的時候你掛了，要你何用？

批處理執(zhí)行是有大數(shù)據(jù)量插入時推薦的做法，使用起來也比較方便。