SQL 连接以包括最左侧表上的所有记录，并且仅显示满足两个条件的两个右侧表中的记录

Question

我需要帮助我的加入和条件。

我想显示hpd_help_desk中的所有记录。

表hpd_associations和chg_infrastructure_change每个hpd_help_desk.INCIDENT_NUMBER有多个记录，因此会返回重复的行。 需要删除重复项。

表hpd_associations仅添加到 chg_infrastructure_change 中的访问记录

删除重复项的建议逻辑：

1. 从表chg_infrastructure_change select 中，只有字段chg_request具有CRQ编号和HPD_ASSOCIATIONS_Request = 17000的行

2. 如果hpd_associations.association_type01每个hpd_help_desk.INCIDENT_NUMBER有多行并且没有chg_request CRQ's / HPD_ASSOCIATIONS_Request = 17000 ，则只需取第一行。

SQL：

SELECT  distinct(hpd_help_desk.INCIDENT_NUMBER),
        
        hpd_associations.association_type01                 as "HPD_ASSOCIATIONS_Request",
        
        chg_infrastructure_change.request_id                as "CHG_Request",
        chg_infrastructure_change.status_reason             as "CHG_Status_Reason",
        chg_infrastructure_change.description2              as "CHG__Desc2",  
        chg_infrastructure_change.infrastructure_change_id  as "CHG_infrastructure_change_id"  
    
FROM    helix_access.hpd_help_desk
        LEFT OUTER JOIN helix_access.hpd_associations ON (hpd_help_desk.incident_number = hpd_associations.request_id02)
        LEFT OUTER JOIN helix_access.chg_infrastructure_change ON (chg_infrastructure_change.infrastructure_change_id  = hpd_associations.request_id01)                                            

WHERE   hpd_associations.association_type01 = 17000 -- 17000 vale = "Caused By" There are 83 values for association_type01

所需的 output 如下所示，包括保留行和远程行的指示：

incident_number hpd_associations_request    chg_request chg_status_reason   chg_infrastructure_change_id
INC000037937498 35000   NULL    NULL    NULL --Keep
INC000037942103 17000   CRQ000000755311 6000    CRQ000000889056 --Keep
INC000037942103 35000   CRQ000000710522 5000    CRQ000000834307 -- Remove
INC000037942103 35000   NULL    NULL    NULL -- Remove
INC000037951750 17000   CRQ000000752354 5000    CRQ000000885032 -- Keep
INC000037984428 35000   NULL    NULL    NULL -- Keep
INC000037985037 35000   NULL    NULL    NULL -- Keep
INC000037985037 30000   NULL    NULL    NULL -- Remove

或者是否应该有一个子查询来从chg_infrastructure_change获取所有记录，其中字段chg_request有一个CRQ号和HPD_ASSOCIATIONS_Request = 17000并进行左连接？

IE

WHERE
.
.
.
        and hpd_help_desk.INCIDENT_NUMBER in (
                                                SELECT  
                                                        
                                                        distinct(hpd_associations.request_id02)
                                                     --   hpd_associations.association_type01                 as "HPD_ASSOCIATIONS_Request",
                                                       -- chg_infrastructure_change.request_id                as "CHG_Request"
                                                        
                                                FROM    helix_access.hpd_associations 
                                                            LEFT OUTER JOIN helix_access.chg_infrastructure_change ON (chg_infrastructure_change.infrastructure_change_id  = hpd_associations.request_id01)                                            
                                                
                                                WHERE   From_unixtime(Cast(hpd_associations.start_date_01 AS BIGINT),'yyyy-MM-dd HH:mm:ss') >= '2022-03-01 00:00:00.000'
                                                        AND hpd_associations.association_type01 = 17000 -- 17000 vale = "Caused By" There are 83 values for association_type01
                                              )          

彼得

Answer 1

在处理外连接时，您需要注意这些连接表中列中出现 NULL 的可能性。 当涉及到where clause时尤其如此，因为如果在引用外部联接表时不允许使用 NULL，则您会创建等同于内部联接的条件。

例如

SELECT
      hpd_help_desk.INCIDENT_NUMBER
    , hpd_associations.association_type01 AS "HPD_ASSOCIATIONS_Request"
    , chg_infrastructure_change.request_id AS "CHG_Request"
    , chg_infrastructure_change.status_reason AS "CHG_Status_Reason"
    , chg_infrastructure_change.description2 AS "CHG__Desc2"
    , chg_infrastructure_change.infrastructure_change_id AS "CHG_infrastructure_change_id"
FROM helix_access.hpd_help_desk
LEFT OUTER JOIN helix_access.hpd_associations
    ON hpd_help_desk.incident_number = hpd_associations.request_id02
LEFT OUTER JOIN helix_access.chg_infrastructure_change
    ON chg_infrastructure_change.infrastructure_change_id = hpd_associations.request_id01
WHERE (hpd_associations.association_type01 = 17000 
         OR hpd_associations.association_type01 IS NULL)

此外，使用 where 子句的替代方法可以扩展join conditions ，例如

LEFT OUTER JOIN helix_access.hpd_associations
    ON hpd_help_desk.incident_number = hpd_associations.request_id02
    AND hpd_associations.association_type01 = 17000
LEFT OUTER JOIN helix_access.chg_infrastructure_change
   ON chg_infrastructure_change.infrastructure_change_id  = hpd_associations.request_id01
   AND chg_infrastructure_change.CRQ IS NOT NULL
   AND chg_infrastructure_change.HPD_ASSOCIATIONS_Request = 17000

这会将结果中包含的行限制为仅满足每个连接中所有条件的行。

关于“如果 hpd_associations.association_type01 每个 hpd_help_desk.INCIDENT_NUMBER 有多行并且没有 chg_request CRQ's / HPD_ASSOCIATIONS_Request = 17000，则只需取第一行。” 这可能需要使用row_number() over()并且这样做需要您加入子查询而不是直接加入表。 例如

LEFT OUTER JOIN (
    SELECT *
    , row_number() over(partition by request_id02
                        order by CRQ) as rn
    FROM helix_access.hpd_associations
    AND hpd_associations.association_type01 = 17000
    ) AS assocs ON hpd_help_desk.incident_number = assocs.request_id02
                AND assocs.rn = 1 /* this is how you get just one row (for each request_id02) */