Question

对于SQL，我相对缺乏经验。我想知道使用CROSS APPLY是否是下面SQL中的最佳选项？

我也认为我在重复工作。这有效，需要几分钟，但我的感觉是可以做得更好。

我正在看的三个主要表格非常大，我们每个人都会看到几百万行。各个插入语句每个都会拉回7-15K行。

DECLARE @Master TABLE
  (
     heyno              NVARCHAR(12),
     postcodestartdttm  DATE,
     postcodeenddttm    DATE,
     lzohistorypostcode NVARCHAR(25),
     biactivitypostcode NVARCHAR(25),
     activityenddttm    DATE
  );

INSERT INTO @Master
SELECT p.pasid,
       par.startdttm,
       par.enddttm,
       pa.postcode AS LZOHistoryPostcode,
       t.postcode  AS BIActivityPostcode,
       t.admitdate
FROM   HEALTHBI.DBO.[lzo_patientaddressrole] AS par
       INNER JOIN HEALTHBI.DBO.[lzo_patientaddress] AS pa
               ON par.[addressoid] = pa.[oid]
                  AND pa.[status] = 'A'
       INNER JOIN HEALTHBI.DBO.lzo_patient AS p
               ON par.identifyingoid = p.oid
       --- 
       CROSS APPLY (SELECT i.admitdate,
                           i.postcode
                    FROM   HEALTHBI_VIEWS.DBO.ip_admission i
                    WHERE  i.patientoid = p.oid
                           AND i.admitdate > par.startdttm
                           AND i.admitdate >= '01 APRIL 2010'
                           AND i.admitdate < ISNULL(par.enddttm, '31-dec-4712')
                           AND par.[status] = 'A'
                           AND par.[identifyingtype] = 'Patient'
                           AND par.[rotypcode] = 'CC_USUALADD'
                           AND pa.[adtypcode] = 'Address'
                           AND pa.postcode <> i.postcode) AS t

INSERT INTO @Master
SELECT p.pasid,
       par.startdttm,
       par.enddttm,
       pa.postcode AS LZOHistoryPostcode,
       t.postcode  AS BIActivityPostcode,
       t.apptstartdate
FROM   HEALTHBI.DBO.[lzo_patientaddressrole] AS par
       INNER JOIN HEALTHBI.DBO.[lzo_patientaddress] AS pa
               ON par.[addressoid] = pa.[oid]
                  AND pa.[status] = 'A'
       INNER JOIN HEALTHBI.DBO.lzo_patient AS p
               ON par.identifyingoid = p.oid
       CROSS APPLY (SELECT i.apptstartdate,
                           i.postcode
                    FROM   HEALTHBI_VIEWS.DBO.op_appointment i
                    WHERE  i.patientoid = p.oid
                           AND i.apptstartdate > par.startdttm
                           AND i.apptstartdate >= '01 APRIL 2010'
                           AND i.apptstartdate < ISNULL(par.enddttm, '31-dec-4712')
                           AND par.[status] = 'A'
                           AND par.[identifyingtype] = 'Patient'
                           AND par.[rotypcode] = 'CC_USUALADD'
                           AND pa.[adtypcode] = 'Address'
                           AND pa.postcode <> i.postcode) AS t

INSERT INTO @Master
SELECT p.pasid,
       par.startdttm,
       par.enddttm,
       pa.postcode AS LZOHistoryPostcode,
       t.postcode  AS BIActivityPostcode,
       t.attenddate
FROM   HEALTHBI.DBO.[lzo_patientaddressrole] AS par
       INNER JOIN HEALTHBI.DBO.[lzo_patientaddress] AS pa
               ON par.[addressoid] = pa.[oid]
                  AND pa.[status] = 'A'
       INNER JOIN HEALTHBI.DBO.lzo_patient AS p
               ON par.identifyingoid = p.oid
       CROSS APPLY (SELECT i.attenddate,
                           i.postcode
                    FROM   HEALTHBI_VIEWS.DBO.ed_attendance i
                    WHERE  i.patientoid = p.oid
                           AND i.attenddate > par.startdttm
                           AND i.attenddate >= '01 APRIL 2010'
                           AND i.attenddate < ISNULL(par.enddttm, '31-dec-4712')
                           AND par.[status] = 'A'
                           AND par.[identifyingtype] = 'Patient'
                           AND par.[rotypcode] = 'CC_USUALADD'
                           AND pa.[adtypcode] = 'Address'
                           AND pa.postcode <> i.postcode) AS t

SELECT m.heyno,
       m.lzohistorypostcode,
       m.biactivitypostcode,
       d.startofweek
FROM   @Master m
       INNER JOIN HEALTHBI_VIEWS.DBO.date_reference AS d
               ON m.activityenddttm = d.datevalue
GROUP  BY m.heyno,
          m.lzohistorypostcode,
          m.biactivitypostcode,
          d.startofweek
ORDER  BY m.heyno

Answer 1

您不想使用CROSS APPLY。这意味着与表值函数一起使用，它有点像光标，一次一条记录，这比设置操作慢得多。

我会尝试使用OUTER JOIN，如下所示。如果不更好地理解数据，很难做到正确，但似乎对于每个患者，您将获得多个入院日期和邮政编码，因此您可能会有一些重复，具体取决于ip_addmission视图返回的内容。如果不需要，您可以考虑使用GROUP BY仅返回最新的录取日期。

此外，避免在WHERE子句中使用函数，因为它们可能导致表扫描。我用下面的OR表达式替换了ISNULL。

SELECT p.pasid,
       par.startdttm,
       par.enddttm,
       pa.postcode AS LZOHistoryPostcode,
       i.postcode  AS BIActivityPostcode,
       i.admitdate
FROM   HEALTHBI.DBO.[lzo_patientaddressrole] AS par
       INNER JOIN HEALTHBI.DBO.[lzo_patientaddress] AS pa
               ON par.[addressoid] = pa.[oid]
                  AND pa.[status] = 'A'
       INNER JOIN HEALTHBI.DBO.lzo_patient AS p
               ON par.identifyingoid = p.oid
       --- 
       LEFT OUTER JOIN HEALTHBI_VIEWS.DBO.ip_admission i
        ON i.patientoid = p.oid

WHERE  i.admitdate > par.startdttm
                           AND i.admitdate >= '01 APRIL 2010'
                           AND (i.admitdate < par.enddttm OR par.enddttm IS NULL)
                           AND par.[status] = 'A'
                           AND par.[identifyingtype] = 'Patient'
                           AND par.[rotypcode] = 'CC_USUALADD'
                           AND pa.[adtypcode] = 'Address'
                           AND pa.postcode <> i.postcode;

CROSS是应用最好的方式还是可能是EXISTS或子查询？

1 个答案: