我正在使用标准SQL在Google Big Query中工作。
我有综合浏览量数据。所有相关的是
| user_id | entity_id | url |
URL的格式为/entities/entity_id/show
或entities/entity_id/reply/new
用户可能会同时使用这两种类型的URL,或者两者都显示,并且可能会重复。
我的目标是一张看起来像这样的桌子
| user_id | entity_id | view_type |
view_type
是“显示”或“新”的地方
每个用户/实体对应该只有一行。如果该view_type
/ user_id
对曾经出现在entity_id
形式的一个网址的旁边,则/entities/entity_id/replies/new
应该是“新的”,但是如果该对没有“新的”,则“显示”网址。如果原始表中没有user_id
/ entity_id
对的示例,则最终表中应该没有它们。
我将在示例中加入with语句,以确保重复性
WITH data AS (
select 1 as user_id, 23 as entity_id, '/entities/23/replies/new' as url
UNION ALL
select 1 as user_id, 23 as entity_id, '/entities/23/show' as url
UNION ALL
select 2 as user_id, 30 as entity_id, '/entities/30/show' as url
)
SELECT * from data
那会建立一个这样的表
| user_id | entity_id | url |
----------------------------------------------------
| 1 | 23 | '/entities/23/replies/new' |
| 1 | 23 | '/entities/23/show' |
| 2 | 30 | '/entities/30/show' |
我可以通过使用两个with
语句对任一类型的网址执行select distinct
,然后重新加入并执行case
语句来实现我的目标,该语句在存在或不存在任何联接的情况下运行为给定的user
/ entity
对工作。
这是我的意思:
WITH data AS (
select 1 as user_id, 23 as entity_id, '/entities/23/replies/new' as url
UNION ALL
select 1 as user_id, 23 as entity_id, '/entities/23/show' as url
UNION ALL
select 2 as user_id, 30 as entity_id, '/entities/30/show' as url
), news AS (
SELECT DISTINCT user_id, entity_id, 1 as found
FROM data
WHERE url like '%new'
), shows AS (
SELECT DISTINCT user_id, entity_id, 1 as found
FROM data
WHERE url like '%show'
)
SELECT DISTINCT d.user_id,
d.entity_id,
CASE WHEN n.found = 1 then 'new'
WHEN s.found = 1 then 'show' end as view_type
FROM data d
LEFT JOIN news n on n.user_id = d.user_id and n.entity_id = d.entity_id
LEFT JOIN shows s on s.user_id = d.user_id and s.entity_id = d.entity_id
显然,样本数据使它看起来比实际的还要令人生畏,但这仍然是一个笨拙的,难以理解的查询,如果我添加另一个我想考虑的view_type
,很难扩展。
我想一定有更好的方法!
在我看来,我可以尝试将user_id
/ entity_id
对的所有url填充到一个数组中,然后使用case
语句对数组进行操作,例如:如果数组中的任何元素都匹配“ new”,则匹配“ new”,等等。)但是我不确定如何进行“元素正则表达式匹配”,或者甚至是可能的。
任何人都能提供的见解,我将不胜感激!
答案 0 :(得分:1)
一种方法是聚合:
SELECT user_id, entity_id,
(CASE WHEN COUNTIF(url like '%new') > 0 THEN 'new' ELSE 'show'
END) as view_type
FROM data
GROUP BY user_id, entity_id