SQL 在时间范围内左加入 Window

Question

我有两个数据集：

dataset_a
time_stamp                   user    group    value
2021-06-20 12:48:24.521         A    video        1
2021-06-15 12:50:24.521         A    video        1
2021-06-10 12:48:24.521         A    video        1    

dataset_b
time_stamp                   user    group    label
2021-06-20 09:40:24.521         A    video       BA
2021-06-19 13:30:24.521         A    video       BB  
2021-06-13 12:48:24.521         A    video       BC  
2021-06-09 12:55:24.521         A    video       BD

我想创建一个数据集，如果数据集 b 在时间戳、用户和组方面与数据集 a 的时间戳相差 1 天以内，则它是一个匹配项。以前有没有人做过类似的事情 left join on dataset_b.timestamp between dataset_a.timestamp and date_add(dataset_a.timestamp,-1)。我希望有灵活性，将来我也可以测试 -7 天，这样它就很容易修改。

预期输出如下：

 dataset_a
time_stamp                   user    group    value    timestamp_b               label
2021-06-20 12:48:24.521         A    video      0.5    2021-06-20 09:40:24.521      BA
2021-06-20 12:48:24.521         A    video      0.5    2021-06-19 13:30:24.521      BB
2021-06-15 12:50:24.521         A    video        1    NULL                       NULL   
2021-06-10 12:48:24.521         A    video        1    2021-06-09 12:55:24.521      BD

Answer 1

JOIN 条件不必只是相等运算符，因此：

SELECT *
FROM dataset_a
LEFT JOIN dataset_b
  ON dataset_b.user = dataset_a.user
 AND dataset_b.group = dataset_a.group
 AND dataset_b.time_stamp BETWEEN dataset_a.time_stamp - INTERVAL '1 day'
                              AND dataset_a.time_stamp ;

是一个有效的连接。

db<>fiddle demo

Answer 2

一些调整...与 Lukasz 一样，您现在可以在雪花中 copy/paste 和运行 :-)

 with dataset_a as (
select '2021-06-20 12:48:24.521'::TIMESTAMP_LTZ time_stamp, 'A' user, 
'video' groups,1 value
union all select '2021-06-15 12:50:24.521'::TIMESTAMP_LTZ time_stamp, 'A' user, 'video' groups,1 value
union all select '2021-06-10 12:48:24.521'::TIMESTAMP_LTZ time_stamp, 'A' user, 'video' groups,1 value 
) , dataset_b as( 
select '2021-06-19 13:30:24.521'::TIMESTAMP_LTZ time_stamp,'A' user,'video' groups,'BB' label  
union all select '2021-06-13 12:48:24.521'::TIMESTAMP_LTZ time_stamp,'A' user,'video' groups,'BC' label  
union all select '2021-06-09 12:55:24.521'::TIMESTAMP_LTZ time_stamp,'A' user,'video' groups,'BD' label  
union all select '2021-06-20 09:40:24.521'::TIMESTAMP_LTZ time_stamp,'A' user,'video' groups,'BA' label) 
SELECT *
FROM dataset_a
LEFT JOIN dataset_b
ON dataset_b.user = dataset_a.user
AND dataset_b.groups = dataset_a.groups
AND dataset_b.time_stamp between dataset_a.time_stamp - INTERVAL '1 day' 
and dataset_a.time_stamp ;

添加了 avg(value) 来清理 dups ... 或者只需在您的密钥上添加窗口化的 avg avg(dataset_a.value) over (partition by dataset_a.time_stamp, dataset_a.user, dataset_a.groups , dataset_b.user)

 with dataset_a as (
 select '2021-06-20 12:48:24.521'::TIMESTAMP_LTZ time_stamp, 'A' user, 
 'video' groups,1 value
 union all select '2021-06-15 12:50:24.521'::TIMESTAMP_LTZ time_stamp, 'A' user, 'video' groups,1 value
 union all select '2021-06-10 12:48:24.521'::TIMESTAMP_LTZ time_stamp, 'A' user, 'video' groups,1 value 
 ) , dataset_b as( 
 select '2021-06-19 13:30:24.521'::TIMESTAMP_LTZ time_stamp,'A' user,'video' groups,'BB' label  
 union all select '2021-06-13 12:48:24.521'::TIMESTAMP_LTZ time_stamp,'A' user,'video' groups,'BC' label  
 union all select '2021-06-09 12:55:24.521'::TIMESTAMP_LTZ time_stamp,'A' user,'video' groups,'BD' label  
 union all select '2021-06-20 09:40:24.521'::TIMESTAMP_LTZ time_stamp,'A' user,'video' groups,'BA' label) 
 SELECT dataset_a.time_stamp, dataset_a.user, dataset_a.groups, avg(dataset_a.value), dataset_b.time_stamp, dataset_b.user, dataset_b.groups,dataset_b.label
 FROM dataset_a
 LEFT JOIN dataset_b
 ON dataset_b.user = dataset_a.user
 AND dataset_b.groups = dataset_a.groups
 AND dataset_b.time_stamp between dataset_a.time_stamp - INTERVAL '1 day' 
 and dataset_a.time_stamp 
 group by 1,2,3,5,6,7,8

SQL 在时间范围内左加入 Window

SQL Left Join Within Timeframe Window

sql

dateadd

left-join

snowflake-cloud-data-platform