Why you need it?

For now. When I execute the sql like

update t1 innert join t2 on t1.id = t2.id set t1.name = 'Mike' where t1.id = 33

and the before image and after image would like this

before image 1 -> select t1.id, t1.name from t1 innert join t2 on t1.id = t2.id where t1.id = ? group by t1.id for update
after image 1 -> select t1.id, t1.name from t1 innert join t2 on t1.id = t2.id where t1.id in (?) group by t1.id

and the time complexity depends on the numbers of join table.

How it could be?

I have a idea that we do not join the table and do not group by the table when we build the image. For example execute the same sql

update t1 innert join t2 on t1.id = t2.id set t1.name = 'Mike' where t1.id = 33

first we found the primary key of t1 and t2

select t1.id, t2.id from t1 left join t2 on t1.id = t2.id where t1.id = 33

second we use the primary key of t1 and t2 to build the image of each table

before image 1 -> select t1.id, t1.name from t1
after image 1 -> select t2.id from t2

Other related information

Add any other context or screenshots about the feature request here.