Question

有两个主要表格。我正在提供必要的专栏。

Designers

Category | deisgnID

Impressions

recId | impressionType | impressionId | impressionAction | session_id

以下是sessions的示例会话。 impressionId是一个多态列（我认为这个术语;我可能错了）包含不同的数据类型和值。

recId   impressionType      impressionId    impressionAction    session_Id
73790   USER                    11182       LOGIN                   acbd1234
73791   UNDOCKED            abcd1234        UNDOCKED            acbd1234
73792   PRODUCT                 1446        TAPPED-WALL     acbd1234
73793   CARTS                   3586        ADDED           acbd1234
73794   CART-PRODUCT            14941       ADDED           acbd1234
73801   PRODUCT                 1465        TAPPED-RECOMMENDATION   acbd1234
73802   CART-PRODUCT            14942       ADDED           acbd1234
73811   PRODUCT                 1465        TAPPED-RECOMMENDATION   acbd1234
73818   PRODUCT                 1446        TAPPED-RECOMMENDATION   acbd1234
73828   PRODUCT                 1965        TAPPED-WALL     acbd1234
73829   CART-PRODUCT           14944        ADDED           acbd1234
73836   PRODUCT                 1952        TAPPED-WALL     acbd1234
73837   CART-PRODUCT            14945       ADDED           acbd1234
73882   PRODUCT                 502         TAPPED-WALL     acbd1234
73883   CART-PRODUCT            14949       ADDED           acbd1234
73897   CART-PRODUCT            14951       ADDED           acbd1234
73942   EMAILED_RECOMM          904         SEND            acbd1234
73943   EMAILED_RECOMM          1586        SEND            acbd1234
73944   EMAIL-NOTIFICATION  abcd@amazon.com SENDMAIL        acbd1234

我使用以下代码合并了两个表

SELECT
         d.category
        ,d.designId
        ,i.session_id
        ,COUNT( IF(i.impressionAction = 'TAPPED'
                OR i.impressionAction = 'TAPPED-LISTPAGE'
                OR i.impressionAction = 'TAPPED-wall'
                OR i.impressionAction = 'TAPPED-RECOMMENDATION') AS SCANS_total
        ,COUNT(IF(i.impressionAction = 'TAPPED', 1, NULL)) AS TAPPED
        ,COUNT(IF(i.impressionAction = 'TAPPED-LISTPAGE', 1, NULL)) AS TAPPED_LISTPAGE
        ,COUNT(IF(i.impressionAction = 'TAPPED-WALL', 1, NULL)) AS TAPPED_WALL
        ,COUNT(IF(i.impressionAction = 'TAPPED-RECOMMENDATION', 1, NULL)) AS TAPPED_RECOMMENDATION
        ,COUNT(IF(i.impressionAction = 'SEND', 1, NULL)) AS IS_ITEM_SENT_BY_EMAIL

    FROM
        Impressions i 
        INNER JOIN designers d 
        ON i.impressionId = d.designId

    WHERE
           i.createDate >= '2014-06-18'
        AND HOUR(i.createDate) >= 10
        AND HOUR(i.createDate) < 21
        AND i.impressionId not like '%amazon.com%'

    GROUP BY
        i.session_id, i.impressionId
    HAVING 
        SCANS_total <> 0


    ORDER BY
        d.category, i.impressionId, SCANS_total desc

基本上，我正在生成按照Category和designId 分解的不同类型的扫描扫描产品的次数 em> session。

我的主要问题是：我无法使用i.impressionId not like '%amazon.com%'查询过滤掉某些电子邮件域名，例如amazon.com 。

对于每个会话，如果用户通过电子邮件发送，impresseiondId下的impressionAction = SENDMAIL和impressionType = EMAIL-NOTIFICATION下都有一个电子邮件地址。

我尝试使用where ... i.impressionId not like '%amazon.com%'从计算中完全过滤掉某些域电子邮件的会话，但这并不起作用。

有没有办法在做我想做的事情时过滤掉某些电子邮件域？

任何想法都将不胜感激！

更新

我今天醒来并意识到解决问题的子查询。这是我写的用于过滤会话的子查询;但是，查询超时并失败。

本质上，查询会生成电子邮件地址属于某个域的所有会话，然后我否定这些会话的存在。 关于如何优化这个并使其工作的任何想法？

where i.session_id NOT IN( SELECT session_id from Impressions where impressionId LIKE '%amazon.com%')

Answer 1

我一直在不遗余力地尝试解决这个问题，尝试提出不同的解决方案（请参阅更新失败的解决方案）。

我想我拥有它。如果有办法改善这一点，请告诉我。

SELECT *
FROM 
(
SELECT
         d.category
        ,d.designId
        ,i.session_id
        ,i.createDate

        ,d.name
        ,d.productTitle
        ,d.colors


        ,COUNT( IF(i.impressionAction = 'TAPPED'
                OR i.impressionAction = 'TAPPED-LISTPAGE'
                OR i.impressionAction = 'TAPPED-wall'
                OR i.impressionAction = 'TAPPED-RECOMMENDATION') AS SCANS_total
        ,COUNT(IF(i.impressionAction = 'TAPPED', 1, NULL)) AS TAPPED
        ,COUNT(IF(i.impressionAction = 'TAPPED-LISTPAGE', 1, NULL)) AS TAPPED_LISTPAGE
        ,COUNT(IF(i.impressionAction = 'TAPPED-WALL', 1, NULL)) AS TAPPED_WALL
        ,COUNT(IF(i.impressionAction = 'TAPPED-RECOMMENDATION', 1, NULL)) AS TAPPED_RECOMMENDATION
        ,COUNT(IF(i.impressionAction = 'SEND', 1, NULL)) AS IS_ITEM_SENT_BY_EMAIL

    FROM
        Impressions i 
        INNER JOIN designers d 
        ON i.impressionId = d.designId

    WHERE
           i.createDate >= '2014-06-18'
        AND HOUR(i.createDate) >= 10
        AND HOUR(i.createDate) < 21

    GROUP BY
        i.session_id, i.impressionId
    HAVING 
        SCANS_total <> 0


    ORDER BY
        d.category, i.impressionId, SCANS_total desc 
    ) AS P
    LEFT JOIN 
    (
            SELECT session_id as sessionsToBeRemoved from Impressions 
            where impressionId LIKE '%amazon.com%'

            GROUP BY session_id
    ) AS U on P.session_id = U.sessionsToBeRemoved 
    WHERE U.sessionsToBeRemoved IS NULL

MySQL从包含不同数据类型的列中过滤某些值

1 个答案: