最早的寫法:

WITH T AS
(SELECT case when Col1 IS NULL OR Col1=N‘‘ then Col2 else Col1 end as Code,case when Col1 IS NULL OR Col1=N‘‘ then 1 else 0 end as Flag FROM YM  WHERE Col_076 BETWEEN ‘2018-07-25‘ AND ‘2018-08-03‘ AND Col_478=N‘xx‘ AND Col_346 LIKE N‘%dd%‘),
D AS (SELECT Code,province,city  FROM Adds)
 
SELECT province AS 省,city as 市,COUNT(1) 票數 FROM 
(SELECT A.DR_250 as province,A.DR_251 as city FROM T INNER JOIN TB AS A ON A.DR_203=T.Code WHERE T.Flag=0
 UNION ALL
 SELECT D.province,D.city FROM T INNER JOIN D ON D.Code=T.Code WHERE T.Flag=1
 UNION ALL
 SELECT ‘‘ AS province,‘‘ AS city FROM T WHERE
 
 Code IS NULL OR Code=N‘‘) AS S 
GROUP BY province,city;

最新的寫法:

SELECT case when Col1 IS NULL OR Col1=N‘‘ then Col2 else Col1 end as Code,case when Col1 IS NULL OR Col1=N‘‘ then 1 else 0 end as Flag into #T FROM YM WHERE Col_076 BETWEEN ‘2018-07-25‘ AND ‘2018-08-03‘ AND Col_478=N‘xx‘ AND Col_346 LIKE
 
 N‘%dd%‘;
SELECT Code,province,city into #D  FROM Adds;
SELECT province AS 省,city as 市,COUNT(1) 票數 FROM 
(SELECT A.DR_250 as province,A.DR_251 as city FROM #T INNER JOIN TB AS A ON A.DR_203=#T.Code WHERE #T.Flag=0
UNION ALL
SELECT #D.province,#D.city FROM #T INNER JOIN #D ON #D.Code=#T.Code WHERE #T.Flag=1
UNION ALL
SELECT ‘‘ AS province,‘‘ AS city FROM #T WHERE Code IS NULL OR Code=N‘‘) AS S GROUP BY province,city;
DROP table #T;
DROP table #D;

新的寫法比原始寫法性能高出太多(原語句執行會超時),最大的原因是對with語句理解有誤!!!

一個SQL語句的優化(sqlserver)