工作中遇到的比较奇怪的一些sql（一些子查询）

在列中进行子查询

1、在一个表中有多个员工ID，比如一个下单员工，一个修改订单的员工，可以使用在列中进行子查询，具体如下：

   (
        SELECT
            staff_name
        FROM
            sp_staff_basic
        WHERE
            sp_staff_basic.id = qc_return_progress_record.created_by
    ) returnProcessName,

    LEFT JOIN sp_staff_basic ON sp_staff_basic.id =         qc_returnrervice.updated_by    

2、left join 自己的表，比如说在一个表中，有一个出发时间字段，又有一个到达时间字段，两个表所对应的节点不一致，但是两个表都记录在订单表中，比如订单的一个状态为出发吗，一个状态为到达，我们需要计算路途中所花的时长,left join 自己的表，就可以取出两个节点所对应的时间，具体如下：

select 
sp_staff_usework.attach_group attachGroup,COUNT(sp_staff_usework.attach_group) countFrequency,sp_staff_basic.id staffId
from (SELECT * from or_order_state_record where or_order_state_record.state_type=3)  n1
LEFT JOIN (SELECT * from or_order_state_record where or_order_state_record.state_type=4) n2  on n1.order_id=n2.order_id
LEFT JOIN sp_staff_basic on sp_staff_basic.id=n1.created_by
LEFT JOIN sp_staff_usework on sp_staff_usework.staff_id=n1.created_by

where TIMESTAMPDIFF(MINUTE,n1.created_date,n2.created_date)>3
GROUP BY sp_staff_usework.attach_group

这种方法还可以用来做统计表，表中有多个统计字段的，比如我下面做的客服的服务质量表：

        SELECT  
        sp_staff_basic.id staffId,
        sp_staff_usework.attach_group attachGroup,
        t1.countFrequency timeoutAudit,
        t2.countFrequency timeOutList,
        t3.countMis countMis,
        t4.countOrderChange countOrderChange,
        t5.countPeople countPeople,
        sp_staff_basic.staff_name staffName,
        sp_staff_usework.position_code positionCode,
        sp_staff_usework.position_class postionClass,
        sp_company.id companyId,
        sp_company.company_name companyName
        from sp_staff_basic
        LEFT JOIN sp_staff_usework 
        on sp_staff_usework.staff_id=sp_staff_basic.id
        LEFT JOIN(
        select 
        sp_staff_usework.attach_group attachGroup,COUNT(sp_staff_usework.attach_group) countFrequency,sp_staff_basic.id staffId
        from (SELECT * from or_order_state_record where or_order_state_record.state_type=3)  n1
        LEFT JOIN (SELECT * from or_order_state_record where or_order_state_record.state_type=4) n2  on n1.order_id=n2.order_id
        LEFT JOIN sp_staff_basic on sp_staff_basic.id=n1.created_by
        LEFT JOIN sp_staff_usework on sp_staff_usework.staff_id=n1.created_by
        <where>
        <if test="param.startCreateDate != null">
        AND n1.created_date <![CDATA[  >= ]]> #{param.startCreateDate,jdbcType=TIMESTAMP}
          </if>
          <if test="param.endCreateDate != null">
        AND n1.created_date <![CDATA[ < ]]> date_add(#{param.endCreateDate,jdbcType=TIMESTAMP},interval 1 day)
          </if>
          and TIMESTAMPDIFF(MINUTE,n1.created_date,n2.created_date)>3
        </where>
        GROUP BY sp_staff_usework.attach_group
        )t1 on sp_staff_usework.attach_group=t1.attachGroup
        LEFT JOIN(
        select 
        sp_staff_usework.attach_group attachGroup,COUNT(sp_staff_usework.attach_group) countFrequency,sp_staff_basic.id staffId
        from (SELECT * from or_order_state_record where or_order_state_record.state_type=0)  n1
        LEFT JOIN (SELECT * from or_order_state_record where or_order_state_record.state_type=3) n2  on n1.order_id=n2.order_id
        LEFT JOIN sp_staff_basic on sp_staff_basic.id=n1.created_by
        LEFT JOIN sp_staff_usework on sp_staff_usework.staff_id=n1.created_by
        <where>
        <if test="param.startCreateDate != null">
        AND n1.created_date <![CDATA[  >= ]]> #{param.startCreateDate,jdbcType=TIMESTAMP}
          </if>
          <if test="param.endCreateDate != null">
        AND n1.created_date <![CDATA[ < ]]> date_add(#{param.endCreateDate,jdbcType=TIMESTAMP},interval 1 day)
          </if>
          and TIMESTAMPDIFF(MINUTE,n1.created_date,n2.created_date)>3
        </where>
        GROUP BY sp_staff_usework.attach_group
        )t2 on sp_staff_usework.attach_group=t2.attachGroup
        LEFT JOIN (
        SELECT COUNT(qc_mistakenum.mistakenum_id) countMis,qc_mistakenum.staff_id staffId,qc_mistakenum.created_date from qc_mistakenum
        <where>
        <if test="param.startCreateDate != null">
        AND qc_mistakenum.created_date <![CDATA[  >= ]]> #{param.startCreateDate,jdbcType=TIMESTAMP}
          </if>
          <if test="param.endCreateDate != null">
        AND qc_mistakenum.created_date <![CDATA[ < ]]> date_add(#{param.endCreateDate,jdbcType=TIMESTAMP},interval 1 day)
          </if>
        </where>
        ) t3 on t3.staffId=sp_staff_basic.id
        LEFT JOIN (
        SELECT count(qc_orderchange.order_id) countOrderChange,qc_orderchange.created_by,qc_orderchange.created_date from qc_orderchange
        <where>
        <if test="param.startCreateDate != null">
        AND qc_orderchange.created_date <![CDATA[  >= ]]> #{param.startCreateDate,jdbcType=TIMESTAMP}
          </if>
          <if test="param.endCreateDate != null">
        AND qc_orderchange.created_date <![CDATA[ < ]]> date_add(#{param.endCreateDate,jdbcType=TIMESTAMP},interval 1 day)
          </if>
        </where>
        ) t4 on t4.created_by=sp_staff_basic.id
        LEFT JOIN (
        select qc_complaint_people.comp_people_id,count(qc_complaint_people.comp_people_id) countPeople,qc_complaint_record.created_date
        from qc_complaint_people 
        LEFT JOIN sp_staff_basic
        on sp_staff_basic.id=qc_complaint_people.comp_people_id
        LEFT JOIN qc_complaint_record
        on qc_complaint_record.comp_id=qc_complaint_people.comp_id
        <where>
        <if test="param.startCreateDate != null">
        AND qc_complaint_record.created_date <![CDATA[  >= ]]> #{param.startCreateDate,jdbcType=TIMESTAMP}
          </if>
          <if test="param.endCreateDate != null">
        AND qc_complaint_record.created_date <![CDATA[ < ]]> date_add(#{param.endCreateDate,jdbcType=TIMESTAMP},interval 1 day)
          </if>
        </where>
        GROUP BY qc_complaint_people.comp_people_id
        ) t5 on t5.comp_people_id=sp_staff_basic.id
        left join sp_company
        on sp_company.id=sp_staff_usework.company_id
        <where>
        <if test="param.companyId != null">
        and sp_company.id = #{param.companyId,jdbcType=VARCHAR}
         </if>
         <if test="param.staffId != null and param.staffId != ''">
        and sp_staff_basic.id = #{param.staffId,jdbcType=INTEGER}
          </if>
          <if test="param.positionCode != null and param.positionCode != ''">
        and sp_staff_usework.position_code = #{param.positionCode,jdbcType=INTEGER}
          </if>
         and (sp_staff_usework.position_code='KHFWZX_KFDDDY' or sp_staff_usework.position_code='KHFWZX_ZJKFDDDY' or
             sp_staff_usework.position_code='KHFWZX_GJKFDDDY' or sp_staff_usework.position_code='ALL_ZGSKFGJJL' or 
             sp_staff_usework.position_code='ALL_ZGSKFJL' or  sp_staff_usework.position_code='KHFWZX_ZSKFDDDY'
        )
        </where>
        ORDER BY sp_staff_basic.id

最后的原型为，具体是使用员工去查询他的职位，通过职位去筛选其他的，具体见上sql脚本：

还有一个为服务满意度的

服务满意度的做了好久，在 AND 与 OR 的合用与业务逻辑上卡了蛮久的时间

首先

我们需要过滤出需要回访的 

需要回访的为：

待回访跟踪中仅包含：电话回访-待回访/未接通；短信回访-超时未提交/不满意-且电话回访为待回访的/非常不满意-且电话回访为待回访的，见列表中的集中组合字段

只有在列表中的字段需要进行回访，其他的直接过滤到完成回访的列表，且还需要过滤点击回访，回访结果为未接通字段

所以在

待回访的sql脚本中我们使用的过滤条件为：

WHERE
    or_task_node.node_type = 8
AND (
    cs_customer.default_return_type IN (0, 1)
    AND qc_returnrervice.sms_return_result IN (2, 7, 9, 3)
    AND qc_returnrervice.phone_return_result IN (2, 3)
    AND (
        qc_return_satisfied_record.return_result IS NULL
        OR qc_return_satisfied_record.return_result = 2
    )
)
GROUP BY
    or_rescue_order.id
ORDER BY
    qc_returnrervice.created_date DESC

回访完成后，需要将其他结果与上面过滤条件过滤掉的字段信息全部进行显示，所以用 AND(A OR B)的结构进行筛选

WHERE
    or_task_node.node_type = 8
AND  (
    (
      qc_returnrervice.sms_return_result NOT IN (2, 7, 9, 3)
        AND qc_returnrervice.phone_return_result NOT IN (2, 3)
    )
    OR (
        qc_returnrervice.sms_return_result IN (2, 7, 9, 3)
        AND qc_returnrervice.phone_return_result IN (2, 3)
        AND qc_return_satisfied_record.return_result IN (0, 1)
    )
)
GROUP BY
    or_rescue_order.id
ORDER BY
    qc_returnrervice.created_date DESC

进行相反的过滤，这样就能展示出来所有的回访信息