MyBatis使用Collection查询多对多或一对多结果集bug - ZimZz

公告

情况描述:当使用JOIN查询,如果SQL查询出来的记录不是按id列排序的,则生成的List结果会有问题

案例:

1) 数据库模型

简而言之一个Goods包含多个Goods_Img

2) Java Bean

Goods.java

public class Goods {

    private Integer goodsId;

    private String goodsName;

    private Integer goodsStorageNum;

    private Integer goodsScore;

    private GoodsStatus goodsStatus;

    private String goodsDescription;

    private List<GoodsImg> goodsImgList;

    ... getter and setter ...

}

GoodImg.java

public class GoodsImg {

    private Integer imgId;

    private Integer goodsId;

    private String imgDir;

    private Integer imgSize;

    private String imgName;

    ... getter and setter ...

}

3) Mapper

    <!-- Result Map -->
    <!-- goods resultmap -->
    <resultMap id="goodsResultMap" type="com.qunar.scoresystem.bean.Goods">
        <id property="goodsId" column="goods_id" />
        <result property="goodsName" column="goods_name" />
        <result property="goodsStorageNum" column="goods_storage_num" />
        <result property="goodsScore" column="goods_score" />
        <result property="goodsDescription" column="goods_description" />
        <result property="goodsStatus" column="goods_status" />
        <collection property="goodsImgList" resultMap="goodsImgResult" />
    </resultMap>

    <!-- goodsimage resultmap -->
    <resultMap id="goodsImgResult" type="com.qunar.scoresystem.bean.GoodsImg">
        <id property="imgId" column="img_id" />
        <result property="goodsId" column="goods_id" />
        <result property="imgDir" column="img_dir" />
        <result property="imgSize" column="img_size" />
        <result property="imgName" column="img_name" />
    </resultMap>

4) 执行的 SQL

select 
    goods.goods_id as goods_id, 
    goods.goods_name as goods_name, 
    goods.goods_storage_num as goods_storage_num, 
    goods.goods_score as goods_score,     
    goods.goods_description as goods_description, 
    goods.goods_status as goods_status , 
    goods_img.img_name as img_name , 
    goods_img.img_dir as img_dir , 
    goods_img.img_size as img_size 
from goods 
join goods_img 
    on goods.goods_id=goods_img.goods_id

5) 结果集

a. 当SQL查询的结果为

　　注意上图中的goods_id顺序为乱序

　　则MyBatis返回的List结果为

Goods{goodsId=1, goodsName='good1', goodsStorageNum=1, goodsScore=1, goodsStatus=[1 | 下架], goodsDescription='1', goodsImgList=[GoodsImg{imgId=null, goodsId=1, imgDir='d1', imgSize=1, imgName='img1'}, GoodsImg{imgId=null, goodsId=1, imgDir='d2', imgSize=2, imgName='img2'}, GoodsImg{imgId=null, goodsId=1, imgDir='d4', imgSize=4, imgName='img4'}, GoodsImg{imgId=null, goodsId=1, imgDir='d6', imgSize=6, imgName='img6'}]}

Goods{goodsId=1, goodsName='good1', goodsStorageNum=1, goodsScore=1, goodsStatus=[1 | 下架], goodsDescription='1', goodsImgList=[GoodsImg{imgId=null, goodsId=1, imgDir='d1', imgSize=1, imgName='img1'}, GoodsImg{imgId=null, goodsId=1, imgDir='d2', imgSize=2, imgName='img2'}, GoodsImg{imgId=null, goodsId=1, imgDir='d4', imgSize=4, imgName='img4'}, GoodsImg{imgId=null, goodsId=1, imgDir='d6', imgSize=6, imgName='img6'}]}

Goods{goodsId=1, goodsName='good1', goodsStorageNum=1, goodsScore=1, goodsStatus=[1 | 下架], goodsDescription='1', goodsImgList=[GoodsImg{imgId=null, goodsId=1, imgDir='d1', imgSize=1, imgName='img1'}, GoodsImg{imgId=null, goodsId=1, imgDir='d2', imgSize=2, imgName='img2'}, GoodsImg{imgId=null, goodsId=1, imgDir='d4', imgSize=4, imgName='img4'}, GoodsImg{imgId=null, goodsId=1, imgDir='d6', imgSize=6, imgName='img6'}]}

　　可见返回的结果中有 三个一模一样的 Goods(id=1,且包含5个GoodsImg), 而我们期待的结果应该是 List{ Goods(id=1), Goods(id=2), Goods(id=3) }

　　b. 当使用的SQL查询结果如下

　　上面的查询结果为id有序结果,正则MyBatis返回的Java结果集为:

Goods{goodsId=1, goodsName='good1', goodsStorageNum=1, goodsScore=1, goodsStatus=[1 | 下架], goodsDescription='1', goodsImgList=[GoodsImg{imgId=null, goodsId=1, imgDir='d1', imgSize=1, imgName='img1'}, GoodsImg{imgId=null, goodsId=1, imgDir='d2', imgSize=2, imgName='img2'}, GoodsImg{imgId=null, goodsId=1, imgDir='d3', imgSize=3, imgName='img3'}, GoodsImg{imgId=null, goodsId=1, imgDir='d4', imgSize=4, imgName='img4'}]}

Goods{goodsId=2, goodsName='good2', goodsStorageNum=2, goodsScore=2, goodsStatus=[1 | 下架], goodsDescription='2', goodsImgList=[GoodsImg{imgId=null, goodsId=2, imgDir='d5', imgSize=5, imgName='img5'}]}

Goods{goodsId=3, goodsName='good3', goodsStorageNum=3, goodsScore=3, goodsStatus=[1 | 下架], goodsDescription='3', goodsImgList=[GoodsImg{imgId=null, goodsId=3, imgDir='d6', imgSize=6, imgName='img6'}]}

　　观察goodsId,我们取得了期待的结果

答案:

　　根据作者本人的解释, MyBatis为了降低内存开销,采用ResultHandler逐行读取的JDBC ResultSet结果集的,这就会造成MyBatis在结果行返回的时候无法判断以后的是否还会有这个id的行返回,所以它采用了一个方法来判断当前id的结果行是否已经读取完成,从而将其加入结果集List,这个方法是:

　　1. 读取当前行记录A,将A加入自定义Cache类,同时读取下一行记录B

　　2. 使用下一行记录B的id列和值为key(这个key由resultMap的<id>标签列定义)去Cache类里获取记录

　　3. 假如使用B的key不能够获取到记录,则说明B的id与A不同,那么A将被加入到List

　　4. 假如使用B的key可以获取到记录,说明A与B的id相同,则会将A与B合并(相当于将两个goodsImg合并到一个List中,而goods本身并不会增加)

　　5. 将B定为当前行,同时读取下一行C,重复1-5,直到没有下一行记录

　　6. 当没有下一行记录的时候,将最后一个合并的resultMap对应的java对象加入到List(最后一个被合并goodsImg的Goods)

所以

a. 当结果行是乱序的,例如BBAB这样的顺序,在记录行A遇到一个id不同的曾经出现过的记录行B时, A将不会被加入到List里(因为Cache里已经存在B的id为key的cahce了)

　　b. 当结果是顺序时,则结果集不会有任何问题,因为记录行 A 不可能遇到一个曾经出现过的记录行B, 所以记录行A不会被忽略,每次遇到新行B时,都不可能使用B的key去Cache里取到值,所以A必然可以被加入到List

在MyBatis中,实现这个逻辑的代码如下

  @Override
  protected void handleRowValues(ResultSet rs, ResultMap resultMap, ResultHandler resultHandler, RowBounds rowBounds, ResultColumnCache resultColumnCache) throws SQLException {
    final DefaultResultContext resultContext = new DefaultResultContext();
    skipRows(rs, rowBounds);
    Object rowValue = null;
    while (shouldProcessMoreRows(rs, resultContext, rowBounds)) {
      final ResultMap discriminatedResultMap = resolveDiscriminatedResultMap(rs, resultMap, null);
      // 下一记录行的id构成的cache key
      final CacheKey rowKey = createRowKey(discriminatedResultMap, rs, null, resultColumnCache);
      Object partialObject = objectCache.get(rowKey);
      // 判断下一记录行是否被记录与cache中,如果不在cache中则将该记录行的对象插入List
      if (partialObject == null && rowValue != null) { // issue #542 delay calling ResultHandler until object 
        if (mappedStatement.isResultOrdered()) objectCache.clear(); // issue #577 clear memory if ordered
        callResultHandler(resultHandler, resultContext, rowValue);
      } 
      // 当前记录行的值
      rowValue = getRowValue(rs, discriminatedResultMap, rowKey, rowKey, null, resultColumnCache, partialObject);
    }
    // 插入最后一记录行的对象到List
    if (rowValue != null) callResultHandler(resultHandler, resultContext, rowValue);
  }