我正处于从Oracle迁移到SQL Datawarehouse Azure的仓库中,并遇到了此查询的问题。
来自Oracle的原始查询 - 它返回1872520
行。
SELECT
*
FROM
STG_REV_APPORTION_CSC_NO t1,
STG_SEP_VL t2,
STG_SEP_VL t3
WHERE
t3.BUSINESS_DATE(+) = t1.BUSINESS_DATE
AND t3.CSC_APP_NO(+) = t1.CSC_APP_NO
AND t3.JOURNEY_NO(+) = t1.JOURNEY_NO
AND t3.PURSE_TXN_CTR(+) = t1.PURSE_TXN_CTR
AND t2.BUSINESS_DATE(+) = t1.BUSINESS_DATE
AND t2.CSC_APP_NO(+) = t1.CSC_APP_NO
AND t2.JOURNEY_NO(+) = t1.JOURNEY_NO
AND
(
t2.TRIP_NO(+) + 1
)
= t1.TRIP_NO
AND
(
t2.MSG_TYPE_CD(+) = 13070
AND t3.MSG_TYPE_CD(+) = 4357
);
从documentation获取线索,我尝试重新写入ANSI:
SELECT COUNT(*)
FROM STG_REV_APPORTION_CSC_NO t1
RIGHT OUTER JOIN STG_SEP_VL t3 ON t3.BUSINESS_DATE = t1.BUSINESS_DATE
AND t3.CSC_APP_NO = t1.CSC_APP_NO
AND t3.JOURNEY_NO = t1.JOURNEY_NO
AND t3.PURSE_TXN_CTR = t1.PURSE_TXN_CTR
RIGHT OUTER JOIN STG_SEP_VL t2 ON t2.BUSINESS_DATE = t1.BUSINESS_DATE
AND t2.CSC_APP_NO = t1.CSC_APP_NO
AND t2.JOURNEY_NO = t1.JOURNEY_NO
AND (t2.TRIP_NO + 1) = t1.TRIP_NO
WHERE t2.MSG_TYPE_CD = 13070 AND t3.MSG_TYPE_CD = 4357
返回零行。 ANSI版本应该在oracle实例上运行 - 它也返回零行。
然后我尝试使用toad上的refactor选项将plus join转换为ANSI。我得到以下
SELECT *
FROM STG_SEP_VL T2
RIGHT OUTER JOIN STG_REV_APPORTION_CSC_NO T1
ON (T2.BUSINESS_DATE = T1.BUSINESS_DATE)
AND (T2.CSC_APP_NO = T1.CSC_APP_NO)
AND (T2.JOURNEY_NO = T1.JOURNEY_NO)
RIGHT OUTER JOIN STG_SEP_VL T3
ON (T3.PURSE_TXN_CTR = T1.PURSE_TXN_CTR)
AND (T3.BUSINESS_DATE = T1.BUSINESS_DATE)
AND (T3.CSC_APP_NO = T1.CSC_APP_NO)
AND (T3.JOURNEY_NO = T1.JOURNEY_NO)
WHERE ( ( (T2.TRIP_NO /*(+)*/
) + 1) = T1.TRIP_NO)
AND ( ( (T2.MSG_TYPE_CD /*(+)*/
) = 13070) AND ( (T3.MSG_TYPE_CD /*(+)*/
) = 4357));
现在,此查询应该在Oracle上运行并返回相同数量的行,然后才能在SQL Server上运行它。但它没有 - 它返回零行。
我错过了什么吗?
答案 0 :(得分:3)
以下是我提出的建议:
SELECT *
FROM stg_rev_apportion_csc_no t1
LEFT JOIN stg_sep_vl t3
ON t1.business_date = t3.business_date AND
t1.csc_app_no = t3.csc_app_no AND
t1.journey_no = t3.journey_no AND
t1.purse_txn_ctr = t3.purse_txn_no AND
4357 = t3.msg_type_cd
LEFT JOIN stg_sep_vl t2
ON t1.business_date = t2.business_date AND
t1.csc_app_no = t2.csc_app_no AND
t1.journey_no = t2.journey_no AND
t1.trip_no = t2.trip_no + 1 AND
13070 = t2.msg_type_cd;
表t2和t3外部连接到t1,因此您要先列出t1并进行左连接,要么首先列出t2和t3并进行右连接。
答案 1 :(得分:2)
如果没有样本数据,很难确定,但我认为where子句应该受到责备。
在where子句中包含t2和t3中的字段会否定外连接的效果,除非您还允许空值(t2.MSG_TYPE_CD = 13070 OR 2.MSG_TYPE_CD IS NULL
)。将这些过滤器移动到连接中可以将不匹配的记录放入结果中。
SELECT
COUNT(*)
FROM
STG_REV_APPORTION_CSC_NO t1
RIGHT OUTER JOIN STG_SEP_VL t3 ON t3.BUSINESS_DATE = t1.BUSINESS_DATE
AND t3.CSC_APP_NO = t1.CSC_APP_NO
AND t3.JOURNEY_NO = t1.JOURNEY_NO
AND t3.PURSE_TXN_CTR = t1.PURSE_TXN_CTR
AND t3.MSG_TYPE_CD = 4357
RIGHT OUTER JOIN STG_SEP_VL t2 ON t2.BUSINESS_DATE = t1.BUSINESS_DATE
AND t2.CSC_APP_NO = t1.CSC_APP_NO
AND t2.JOURNEY_NO = t1.JOURNEY_NO
AND (t2.TRIP_NO + 1) = t1.TRIP_NO
AND t2.MSG_TYPE_CD = 13070
;
我并非100%确信此查询是正确的。我怀疑右外连接应该用左外连接替换。这将返回t1中的每条记录,只返回匹配的t2和t3的记录。
答案 2 :(得分:1)
很难找到导致这种不匹配的确切原因,但我认为您已经互换了表STG_SEP_VL中PURSE_TXN_CTR列的连接条件。
SELECT *
FROM STG_REV_APPORTION_CSC_NO t1
RIGHT
JOIN STG_SEP_VL t2
ON t2.BUSINESS_DATE = t1.BUSINESS_DATE
AND t2.CSC_APP_NO = t1.CSC_APP_NO
AND t2.JOURNEY_NO = t1.JOURNEY_NO
AND ( t2.TRIP_NO + 1 ) = t1.TRIP_NO
RIGHT
JOIN STG_SEP_VL t3
ON t3.BUSINESS_DATE = t1.BUSINESS_DATE
AND t3.CSC_APP_NO = t1.CSC_APP_NO
AND t3.JOURNEY_NO = t1.JOURNEY_NO
AND t3.PURSE_TXN_CTR = t1.PURSE_TXN_CTR
WHERE ( t2.MSG_TYPE_CD = 13070 AND t3.MSG_TYPE_CD = 4357 );