Question

我有一张看起来像这样的桌子

visit_id    event_list    seq_num    search_term
    1        search          1         sample
    1        click           3         null
    2        search          1         sample
    3        search          1         sample
    3        click           2         null

我想在一天结束时

search_term    searches    visits    product_clicks
sample             3         3             1

请记住，只有在搜索后直接点击它。

现在，我可以进行正确的访问和搜索，直到增加产品点击次数为止。我通过单击表重新加入表，然后在visit_id上加入。当我加入此连接时，我的搜索量猛增了很多，我不知道为什么。

这是我的代码

select a.search_term
,count(a.visit_id) AS searches
,count(distinct a.visit_id) visits
,count(case when cast(b.seq_num as int) = cast(a.seq_num as int) + 1 then a.visit_id end) prod_click
from Table a join (select visit_id
                    ,seq_num
                    ,event_list
                    from Table
                    where event_list = 'click'
                    and seq_num != 'null') b
                    on b.visit_id = a.visit_id
where event_list = 'search'
group by 1

编辑：查询

select a.search_term
,count(distinct a.visit_id) as visits
,count(distinct case when a.search_term != 'null' then a.visit_id end) as searches
,count(distinct case when b.seq_num = a.seq_num + 1 then 1 end) as prod_click
from T a, T b

可以满足我的需求，但是我正在使用的表非常庞大，因此无法使用。有谁有办法使它变得更好？

Answer 1

您可以在第二行代码中进行“案例分析”

count(distinct case when event_list = "click" then a.visit_id else 0) AS product_clicks,
count(distinct case when event_list = "searches" then a.visit_id else 0) AS searches,

我猜您不需要加入表格。谢谢！

Answer 2

您可以尝试使用条件汇总函数来实现。

SUM和CASE WHEN

CREATE TABLE T(
  visit_id int,
  event_list varchar(50),
  seq_num int,
  search_term varchar(50)
);



insert into t values (1,'search',1,'sample');
insert into t values (1,'click',2,'sample');
insert into t values (2,'search',1,'sample');
insert into t values (3,'search',1,'sample');
insert into t values (3,'click',2,'sample');

查询1 ：

select search_term,
       COUNT(distinct visit_id) searches,
       SUM(CASE WHEN event_list = 'search' THEN 1 ELSE 0 END) visits,
       SUM(CASE WHEN event_list = 'click' THEN 1 ELSE 0 END) product_clicks
FROM T
GROUP BY search_term

Results ：

| search_term | searches | visits | product_clicks |
|-------------|----------|--------|----------------|
|      sample |        3 |      3 |              2 |

编辑

您可以尝试将LEAD函数用于下一个状态，并比较seq_num到ROW_NUMBER是连续的

CREATE TABLE T(
  visit_id int,
  event_list varchar(50),
  seq_num int,
  search_term varchar(50)
);



insert into t values (1,'search',1,'sample');
insert into t values (1,'click', 3,null);
insert into t values (2,'search',1,'sample');
insert into t values (3,'search',1,'sample');
insert into t values (3,'click', 2,null);

查询1 ：

with cte as (
  select *,
      (ROW_NUMBER() OVER(order by visit_id) - seq_num) rn
  FROM T
), cte1 as (
  select *,LEAD(event_list) OVER(PARTITION BY visit_id,rn order by visit_id) nextEvent 
  from cte 
)
SELECT * FROM (
  SELECT search_term,       
         COUNT(distinct visit_id) searches,
         SUM(CASE WHEN event_list = 'search' THEN 1 ELSE 0 END) visits,
         COUNT(CASE WHEN event_list = 'search' and nextEvent = 'click' then 1 end) product_clicks
  FROM cte1 
  group by search_term
)t
where search_term is not null

Results ：

| search_term | searches | visits | product_clicks |
|-------------|----------|--------|----------------|
|      sample |        3 |      3 |              1 |

SQL连接更改计数

2 个答案: