如何在 HIVEQL 中的特定列值上进行 LEFT OUTER JOIN？答案

【问题标题】：How to LEFT OUTER JOIN on specific column values in HIVEQL?如何在 HIVEQL 中的特定列值上进行 LEFT OUTER JOIN？
【发布时间】：2018-03-28 17:21:52
【问题描述】：

当我运行以下 HiveQL 代码时出现错误：执行错误：编译语句时出错：失败：SemanticException [错误 10004]：第 1 行：2112 无效的表别名或列引用“T3”

    SELECT *        
    FROM CC_CLAIM_EXT T1
    INNER JOIN CC_EXPOSURE_EXT T2 ON (T1.ID = T2.CLAIMID)
    LEFT OUTER JOIN CC_POLICY_EXT T3 ON (T1.POLICYID = T3.ID)
    LEFT OUTER JOIN CC_COVERAGE_EXT T4 ON (T2.COVERAGEID = T4.ID)
    LEFT OUTER JOIN CC_TRANSACTION_EXT T5 ON (T2.ID = T5.EXPOSUREID)
    LEFT OUTER JOIN CC_TRANSACTIONSET_EXT T6 ON (T5.TRANSACTIONSETID = T6.ID)
    LEFT OUTER JOIN CC_TRANSACTIONLINEITEM_EXT T7 ON (T5.ID = T7.TRANSACTIONID)
    LEFT OUTER JOIN CC_RISKUNIT_EXT T12 ON (T4.RISKUNITID = T12.ID)
    LEFT OUTER JOIN CC_CLASSCODE_EXT T13 ON (T12.CLASSCODEID = T13.ID)

    LEFT OUTER JOIN (SELECT TT12.CLAIMID
                            ,CASE WHEN COUNT(TT13.PRIMARYBODYPART) > 1 THEN 10010 ELSE MAX(TT13.PRIMARYBODYPART) END AS PRIMARYBODYPART
                            ,CASE WHEN COUNT(TT13.DETAILEDBODYPART) > 1 THEN 10010 ELSE MAX(TT13.DETAILEDBODYPART) END AS DETAILEDBODYPART
                    FROM CC_INCIDENT_EXT TT12
                    LEFT OUTER JOIN CC_BODYPART_EXT TT13 ON (TT12.ID = TT13.INCIDENTID)
                    GROUP BY TT12.CLAIMID) T14 
    ON (T1.ID = T14.CLAIMID AND T3.POLICYTYPE IN(10022,10023))

    WHERE T1.STATE IN(2,3)
        AND T2.STATE IN(2,3)
        AND T6.APPROVALSTATUS = 1
        AND T7.RETIRED = 0

    ORDER BY CLAIMNUMBER
        ,EXPOSUREID
        ,TRANSACTIONID

我已经把范围缩小到了：

    ON (T1.ID = T14.CLAIMID AND T3.POLICYTYPE IN(10022,10023))

如果我删除：

    AND T3.POLICYTYPE IN(10022,10023)

代码运行良好。有没有更好的方法来限制 HiveQL 中的这种联接？

【问题讨论】：

标签： hadoop hive hiveql

【解决方案1】：

错误是因为您在 T1 和 T14 之间的左连接上放置了一个引用，但是您将 T3 放在了条件上。要将您的查询限制在您指定的 T3.policy_id，您应该沿着您离开连接 T1 和 T3 的行输入。见下文第 4 行：

SELECT *        
    FROM CC_CLAIM_EXT T1
    INNER JOIN CC_EXPOSURE_EXT T2 ON (T1.ID = T2.CLAIMID)
    LEFT OUTER JOIN CC_POLICY_EXT T3 ON (T1.POLICYID = T3.ID AND T3.POLICYTYPE IN(10022,10023))
    LEFT OUTER JOIN CC_COVERAGE_EXT T4 ON (T2.COVERAGEID = T4.ID)
    LEFT OUTER JOIN CC_TRANSACTION_EXT T5 ON (T2.ID = T5.EXPOSUREID)
    LEFT OUTER JOIN CC_TRANSACTIONSET_EXT T6 ON (T5.TRANSACTIONSETID = T6.ID)
    LEFT OUTER JOIN CC_TRANSACTIONLINEITEM_EXT T7 ON (T5.ID = T7.TRANSACTIONID)
    LEFT OUTER JOIN CC_RISKUNIT_EXT T12 ON (T4.RISKUNITID = T12.ID)
    LEFT OUTER JOIN CC_CLASSCODE_EXT T13 ON (T12.CLASSCODEID = T13.ID)
    LEFT OUTER JOIN (SELECT TT12.CLAIMID
                            ,CASE WHEN COUNT(TT13.PRIMARYBODYPART) > 1 THEN 10010 ELSE MAX(TT13.PRIMARYBODYPART) END AS PRIMARYBODYPART
                            ,CASE WHEN COUNT(TT13.DETAILEDBODYPART) > 1 THEN 10010 ELSE MAX(TT13.DETAILEDBODYPART) END AS DETAILEDBODYPART
                    FROM CC_INCIDENT_EXT TT12
                    LEFT OUTER JOIN CC_BODYPART_EXT TT13 ON (TT12.ID = TT13.INCIDENTID)
                    GROUP BY TT12.CLAIMID) T14 
    ON T1.ID = T14.CLAIMID
    WHERE T1.STATE IN(2,3)
        AND T2.STATE IN(2,3)
        AND T6.APPROVALSTATUS = 1
        AND T7.RETIRED = 0
    ORDER BY CLAIMNUMBER
        ,EXPOSUREID
        ,TRANSACTIONID

【讨论】：