Help Center/ Data Lake Insight/ FAQs/ Problems Related to SQL Jobs/ O&M Guide/ The on Clause Is Not Added When Tables Are Joined. Cartesian Product Query Causes High Resource Usage of the Queue, and the Job Fails to Be Executed
Updated on 2022-11-09 GMT+08:00
View PDF
Share

The on Clause Is Not Added When Tables Are Joined. Cartesian Product Query Causes High Resource Usage of the Queue, and the Job Fails to Be Executed

Symptom

The on clause was not added to the SQL statement for joining tables. As a result, the Cartesian product query occurs due to multi-table association, and the queue resources were used up. Job execution fails on the queue.

For example, the following SQL statement left-joins three tables without the on clause. 
select 
     case 
        when to_char(from_unixtime(fs.special_start_time), 'yyyy-mm-dd') < '2018-10-12' and row_number() over(partition by fg.goods_no order by fs.special_start_time asc) = 1 then 1
        when to_char(from_unixtime(fs.special_start_time), 'yyyy-mm-dd') >= '2018-10-12' and fge.is_new = 1 then 1 
        else 0 end as is_new
from testdb.table1 fg 
left join testdb.table2 fs
left join testdb.table3  fge
where to_char(from_unixtime(fs.special_start_time), 'yyyymmdd') = substr('20220601',1,8)

Solution

When you use join to perform multi-table query, you must use the on clause to reduce the data volume.

The following example uses the on clause for the table join, which greatly reduces the result set of associated query and improves the query efficiency. 
select 
     case 
        when to_char(from_unixtime(fs.special_start_time), 'yyyy-mm-dd') < '2018-10-12' and row_number() over(partition by fg.goods_no order by fs.special_start_time asc) = 1 then 1
        when to_char(from_unixtime(fs.special_start_time), 'yyyy-mm-dd') >= '2018-10-12' and fge.is_new = 1 then 1 
        else 0 end as is_new
from testdb.table1 fg 
left join testdb.table2 fs on fg.col1 = fs.col2
left join testdb.table3  fge on fg.col3 = fge.col4
where to_char(from_unixtime(fs.special_start_time), 'yyyymmdd') = substr('20220601',1,8)

Parent topic: O&M Guide