使用不包含在GROUP BY中的字段的IF语句

Question

我正在尝试构建一个表，每次访问我的网站时会有一行，然后根据用户是否执行某些操作来显示0或1的列。

为了构建这些列，我使用了IF语句，但这样做意味着我必须将IF语句中引用的字段添加到GROUP BY列表中，这反过来意味着我现在每次访问都有多行，如图所示下面：

如何调整我的代码，以便每visit_ref只有一行，显示该会话中的所有互动？

这是我的代码：

SELECT CONCAT(CAST(fullVisitorId AS STRING),CAST(visitId AS STRING)) AS visit_ref,
customDimension.value AS UserID,
# Event actions
IF(hits.eventInfo.eventAction LIKE "Basket Icon Click",1,0) AS basket_icon_click,
IF(hits.eventInfo.eventAction LIKE "Add to Basket PLP",1,0) AS basket_add_plp,
# Event labels
IF(hits.eventInfo.eventLabel LIKE "Close Mini Basket",1,0) AS closed_minibasket_click,
IF(hits.eventInfo.eventLabel LIKE "%Delivery",1,0) AS checked_delivery_pdp,
IF(hits.eventInfo.eventLabel LIKE "%Returns",1,0) AS checked_returns_pdp,
# Page interactions
IF(hits.page.pagepath LIKE "%/Account/SignIn%",1,0) AS sign_in,
# Device category
IF(device.deviceCategory LIKE "mobile",1,0) AS device_mobile,
IF(device.deviceCategory LIKE "desktop",1,0) AS device_computer,
IF(device.deviceCategory LIKE "tablet",1,0) AS device_tablet,

IF(COUNT(DISTINCT hits.transaction.transactionid)>0,1,0) AS ordered
    FROM `PROJECT.DATASET.ga_sessions_20*` AS t
      CROSS JOIN UNNEST(hits) AS hits
      CROSS JOIN UNNEST(t.customdimensions) AS customDimension
    WHERE parse_date('%y%m%d', _table_suffix) between 
    DATE_sub(current_date(), interval 1 day) and
    DATE_sub(current_date(), interval 1 day)
    AND customDimension.index = 2
    GROUP BY visit_ref, UserID, device.deviceCategory, hits.eventInfo.eventAction, hits.eventInfo.eventCategory, hits.eventInfo.eventLabel, hits.page.pagepath, geoNetwork.country

Answer 1

使用条件聚合。 。 。 case而不是if()：

SELECT CONCAT(CAST(fullVisitorId AS STRING), CAST(visitId AS STRING)) AS visit_ref,
        customDimension.value AS UserID,
# Event actions
        max(case when hits.eventInfo.eventAction LIKE 'Basket Icon Click' then 1 else 0 end) AS basket_icon_click,
        max((case when hits.eventInfo.eventAction LIKE 'Add to Basket PLP' then 1 else 0 end) AS basket_add_plp,
        . . .
from . . .
group by visit_ref, user_id;