Hive - SUM 中的 ELSE答案

【问题标题】：Hive - ELSE in SUMHive - SUM 中的 ELSE
【发布时间】：2015-07-18 07:54:20
【问题描述】：

以下内容在 RDBMS 数据库上作为 SQL 运行良好。但是，这无法在 hive 上运行。

SELECT
        entry_date,
        customer,
        cust_loc,
        SUM(run_time) AS TOTAL_RUN,
        SUM(CASE WHEN run_time BETWEEN 10000 AND 20000 THEN entry_date ELSE 0 END) AS SLOT_1,
        SUM(CASE WHEN run_time BETWEEN 200001 AND 30000 THEN entry_date ELSE 0 END) AS SLOT_2,
        SUM(CASE WHEN run_time BETWEEN 31000 AND 40000 THEN entry_date ELSE 0 END) AS SLOT_3
FROM
        cust_run_details
WHERE
        app_env IN ('a','b')
        AND entry_date = '2015-02-01'
        AND flag_set='U'
GROUP BY
        customer,
        customer,
        cust_loc

我在 HIVE 上运行时遇到的错误是

FAILED: SemanticException [Error 10016]: Line 6:80 Argument type mismatch '0': The expression after ELSE should have the same type as those after THEN: "string" is expected but "int" is found

如何将其转换为 Hive 等效项？请任何想法在这里对我有帮助

【问题讨论】：

将 SUM 更改为 COUNT 并去掉“ELSE 0”。花了很长时间才到那里。叹息..

标签： sql hadoop hive bigdata

【解决方案1】：

假设entry_date 是date，您可能不想sum() 它。如果您想要不同范围内的运行时间总和，那么您需要这样的select：

 SELECT entry_date,
        customer,
        cust_loc,
        SUM(run_time) AS TOTAL_RUN,
        SUM(CASE WHEN run_time BETWEEN 10000 AND 20000 THEN run_time ELSE 0 END) AS SLOT_1,
        SUM(CASE WHEN run_time BETWEEN 200001 AND 30000 THEN run_time ELSE 0 END) AS SLOT_2,
        SUM(CASE WHEN run_time BETWEEN 31000 AND 40000 THEN run_time ELSE 0 END) AS SLOT_3

如果您想要计算不同范围内的运行时间，那么：

 SELECT entry_date,
        customer,
        cust_loc,
        SUM(run_time) AS TOTAL_RUN,
        SUM(CASE WHEN run_time BETWEEN 10000 AND 20000 THEN 1 ELSE 0 END) AS SLOT_1,
        SUM(CASE WHEN run_time BETWEEN 200001 AND 30000 THEN 1 ELSE 0 END) AS SLOT_2,
        SUM(CASE WHEN run_time BETWEEN 31000 AND 40000 THEN 1 ELSE 0 END) AS SLOT_3

如果entry_date 可能是NULL，并且您想忽略这些记录，那么您可以在WHEN 条件中包含and entry_date is not null。

【讨论】：

谢谢戈登。感谢您的帮助。

【解决方案2】：

将 SUM 更改为 COUNT 并去掉“ELSE 0”。花了很长时间才到那里。叹息..

SELECT
        entry_date,
        customer,
        cust_loc,
        SUM(run_time) AS TOTAL_RUN,
        COUNT(CASE WHEN run_time BETWEEN 10000 AND 20000 THEN entry_date END) AS SLOT_1,
        COUNT(CASE WHEN run_time BETWEEN 200001 AND 30000 THEN entry_date END) AS SLOT_2,
        COUNT(CASE WHEN run_time BETWEEN 31000 AND 40000 THEN entry_date END) AS SLOT_3
FROM
        cust_run_details
WHERE
        app_env IN ('a','b')
        AND entry_date = '2015-02-01'
        AND flag_set='U'
GROUP BY
        customer,
        customer,
        cust_loc

【讨论】：