Sybase SQL - 从查询结果中删除“半重复”答案

【问题标题】：Sybase SQL - Remove "semi-duplicates" from query resultsSybase SQL - 从查询结果中删除“半重复”
【发布时间】：2013-03-21 15:50:04
【问题描述】：

我有一个查询使用两个SELECT 语句，它们使用UNION ALL 组合在一起。这两个语句都从相似的表中提取数据以填充查询结果。我正在尝试从查询中删除“半重复”行，但这样做时遇到问题。

我的查询如下：

    SELECT DISTINCT * 
FROM
    (
    SELECT
        TeamNum = CASE 
                WHEN T.TeamName = 'Alpha Team'
                    THEN '1'
                WHEN T.TeamName IN ('Bravo Team', 'Charlie Team')
                    THEN '2'
                WHEN T.TeamName = 'Delta Team'
                    THEN '3'
                ELSE '<Undefined>'
                END,
        P.PatientLastName AS LastName,
        P.PatientFirstName AS FirstName,
        R.PrimaryCity AS City,
        ReimbursorName = CASE
                WHEN RE.ReimbursorDescription = 'Medicare'
                    Then 'R1'
                WHEN RE.ReimbursorDescription = 'Medicaid'
                    Then 'R2'
                ELSE 'R3'
                END,
        P.PatientID AS PatientID
    FROM 
        PatReferrals PR LEFT JOIN Patient P ON PR.PatientID = P.PatientID,
        Patient P LEFT OUTER JOIN Rolodex R ON P.RolodexID = R.RolodexID,
        PatReferrals PR LEFT OUTER JOIN PatReimbursors PRE ON PR.PatientID = PRE.PatientID,
        PatReimbursors PRE LEFT OUTER JOIN Reimbursors RE ON PRE.ReimbursorID = RE.ReimbursorID,
        PatReferrals PR FULL OUTER JOIN Teams T ON PR.TeamID = T.TeamID,
    WHERE 
        PR.ReferralDate BETWEEN GETDATE()-4 AND GETDATE()-1
        AND PR.Status <> 'R' 
        AND PRE.CoveragePriority = '1'
        AND PRE.ExpirationDate IS NULL 

    UNION ALL

    SELECT
        TeamNum = CASE 
                WHEN T.TeamName = 'Alpha Team'
                    THEN '1'
                WHEN T.TeamName IN ('Bravo Team', 'Charlie Team')
                    THEN '2'
                WHEN T.TeamName = 'Delta Team'
                    THEN '3'
                ELSE '<Undefined>'
                END,
        P.PatientLastName AS LastName,
        P.PatientFirstName AS FirstName,
        R.PrimaryCity AS City,
        ReimbursorName = CASE
                WHEN RE.ReimbursorDescription = 'Medicare'
                    Then 'E1'
                WHEN RE.ReimbursorDescription = 'Medicaid'
                    Then 'E2'
                ELSE 'E3'
                END,
        P.PatientID AS PatientID
    FROM 
        PatReferrals PR LEFT JOIN Patient P ON PR.PatientID = P.PatientID,
        Patient P LEFT OUTER JOIN Rolodex R ON P.RolodexID = R.RolodexID,
        PatReferrals PR LEFT OUTER JOIN PatEligibilities PE ON PR.PatientID = PE.PatientID,
        PatEligibilities PE LEFT OUTER JOIN Reimbursors RE ON PE.ReimbursorID = RE.ReimbursorID,
        PatReferrals PR FULL OUTER JOIN Teams T ON PR.TeamID = T.TeamID,
    WHERE 
        PR.ReferralDate BETWEEN GETDATE()-4 AND GETDATE()-1
        AND PR.Status <> 'R' 
        AND PE.Status <> 'V'
        AND PE.ApplicationDate BETWEEN DATE(PR.ReferralDate)-5 AND DATE('2100/01/01')
    ) 

AS DUMMYTBL

ORDER BY 
    DUMMYTBL.LastName ASC,
    DUMMYTBL.FirstName ASC

我在运行查询时收到的结果如下：

3   Doe Jane    Town    R1  19874
1   Roe John    City    R3  50016
1   Roe John    City    E1  50016
2   Smith   Jane    Town    E3  33975

一旦从原始查询中引入结果，我需要删除的数据是基于特定条件的重复行。每个人只能列出一次，并且他们必须有一个支付来源（R1、R2、R3、E1、E2、E3）。如果有 R#，则不能为该人列出 E#。如果没有 R#，则必须列出 E#。如我的示例结果所示，第 2 行和第 3 行列出了同一个人，但有两个支付来源（R3 和 E1）。

如何使用我列出的标准让每个人只显示一行？

编辑：修改了 SQL 查询以显示来自 WHERE 子句的原始变量，以便显示查询的更多详细信息。 PatReimbursors 和 PatEligibilities 表具有相似的数据，但为了提取正确的数据，标准不同。

【问题讨论】：

TeamNum = T.TeamName,?是不是打错字了？
这真的是一个有效的查询吗？ from 子句似乎多次出现相同的别名 - 看起来好像永远无法运行。另外，如果有多个“R”源，哪个优先？
@Rachcha - 我缩短了查询的那部分，因为它最初是一个 CASE 语句以从 TeamName 返回数字。 @Mark Bannister - 是的，这个查询确实有效。

标签： sql sybase duplicate-removal sybase-asa

【解决方案1】：

您的查询没有意义。我将首先消除由, 在from 子句中生成的隐式笛卡尔积。

我的猜测是 from 子句应该是：

FROM 
    PatReferrals PR LEFT JOIN
    Patient P
    ON PR.PatientID = P.PatientID left outer join
    Rolodex R
    ON P.RolodexID = R.RolodexID left outer join
    PatEligibilities PE
    ON PR.PatientID = PE.PatientID left outer join
    Reimbursors RE
    ON PE.ReimbursorID = RE.ReimbursorID left outer join
    Teams T ON PR.TeamID = T.TeamID

执行此操作后，您可能不需要union all 或select distinct。您可以在同一个查询中同时输入报销金额和资格。

【讨论】：

PatReferrals 表分叉给 PatReimbursors 和 PatEligibilities。为了让两个值都填充到同一个查询中，必须插入两个相似的 SELECT 语句的 UNION。没有它，我要么从 PatReimbursors 要么从 PatEligibilities 获取数据，但不能同时从两者中获取。
没关系。这可以防止一种类型的笛卡尔积。但是，您通过在from 子句中包含“，”来介绍多个其他人。

【解决方案2】：

使用一个或多个子查询。

应使用以下模式编写整体查询：

 Select Distinct [Person Data]
 From PersonTable
     left Join to otherTable1 -- add outer join for each table you need data from
         On [Conditions that ensure join can generate only one row per person,
               ... and specify which of possibly many rows to get...]

确保条件消除了连接从另一个 [outer] 表中的 person 表中的每个 person 行生成多个行的任何可能性。这可能（并且经常）要求连接条件基于子查询，例如...

 Select Distinct [Person Data]
 From PersonTable p
     left Join to employments e -- add outer join for each table you need data from
         On e.PersonId = p.PersonId 
            and e.HireDate = (Select Max(hiredate) from employments
                              where personId = p.PersonId)

【讨论】：

【解决方案3】：

在今天处理了相当长的一段时间后，我找到了解决我遇到的问题的方法。这是有效的解决方案，并提取了我需要的正确信息：

SELECT DISTINCT 
    TeamNum,
    LastName,
    FirstName,
    City,
    ReimbursorName = CASE
        WHEN max(ReimbursorName) IN ('R1', 'E1')
            THEN '1' 
        WHEN max(ReimbursorName) IN ('R2', 'E2')
            THEN '2'
        ELSE '3'
        END,
    PatientID 
FROM
    (
    SELECT
        TeamNum = CASE 
                WHEN T.TeamName = 'Alpha Team'
                    THEN '1'
                WHEN T.TeamName IN ('Bravo Team', 'Charlie Team')
                    THEN '2'
                WHEN T.TeamName = 'Delta Team'
                    THEN '3'
                ELSE '<Undefined>'
                END,
        P.PatientLastName AS LastName,
        P.PatientFirstName AS FirstName,
        R.PrimaryCity AS City,
        ReimbursorName = CASE
                WHEN RE.ReimbursorDescription = 'Medicare'
                    Then 'R1'
                WHEN RE.ReimbursorDescription = 'Medicaid'
                    Then 'R2'
                ELSE 'R3'
                END,
        P.PatientID AS PatientID
    FROM 
        PatReferrals PR LEFT JOIN Patient P ON PR.PatientID = P.PatientID,
        Patient P LEFT OUTER JOIN Rolodex R ON P.RolodexID = R.RolodexID,
        PatReferrals PR LEFT OUTER JOIN PatReimbursors PRE ON PR.PatientID = PRE.PatientID,
        PatReimbursors PRE LEFT OUTER JOIN Reimbursors RE ON PRE.ReimbursorID = RE.ReimbursorID,
        PatReferrals PR FULL OUTER JOIN Teams T ON PR.TeamID = T.TeamID
    WHERE 
        PR.ReferralDate BETWEEN GETDATE()-4 AND GETDATE()-1
        AND PR.Status <> 'R' 
        AND PRE.CoveragePriority = '1'
        AND PRE.ExpirationDate IS NULL 

    UNION ALL

    SELECT
        TeamNum = CASE 
                WHEN T.TeamName = 'Alpha Team'
                    THEN '1'
                WHEN T.TeamName IN ('Bravo Team', 'Charlie Team')
                    THEN '2'
                WHEN T.TeamName = 'Delta Team'
                    THEN '3'
                ELSE '<Undefined>'
                END,
        P.PatientLastName AS LastName,
        P.PatientFirstName AS FirstName,
        R.PrimaryCity AS City,
        ReimbursorName = CASE
                WHEN RE.ReimbursorDescription = 'Medicare'
                    Then 'E1'
                WHEN RE.ReimbursorDescription = 'Medicaid'
                    Then 'E2'
                ELSE 'E3'
                END,
        P.PatientID AS PatientID
    FROM 
        PatReferrals PR LEFT JOIN Patient P ON PR.PatientID = P.PatientID,
        Patient P LEFT OUTER JOIN Rolodex R ON P.RolodexID = R.RolodexID,
        PatReferrals PR LEFT OUTER JOIN PatEligibilities PE ON PR.PatientID = PE.PatientID,
        PatEligibilities PE LEFT OUTER JOIN Reimbursors RE ON PE.ReimbursorID = RE.ReimbursorID,
        PatReferrals PR FULL OUTER JOIN Teams T ON PR.TeamID = T.TeamID
    WHERE 
        PR.ReferralDate BETWEEN GETDATE()-4 AND GETDATE()-1
        AND PR.Status <> 'R' 
        AND PE.Status <> 'V'
        AND PE.ApplicationDate BETWEEN DATE(PR.ReferralDate)-5 AND DATE('2100/01/01')
    ) 

AS DUMMYTBL
GROUP BY
    TeamNum,
    LastName,
    FirstName,
    City,
    PatientID
ORDER BY 
    DUMMYTBL.LastName ASC,
    DUMMYTBL.FirstName ASC

感谢您提供的所有回复。

【讨论】：