具有多种连接可能性的连接的 MySQL 更新

Question

概念怀疑：

我已经尝试了很多东西（MySQL），但这个更新的唯一工作方式就是我在这里描述的方式。它正在工作，但我想确认一下：

当我们在更新中使用join时，我将设置的值来自正在加入的表，并且有多个join对于每一行的可能性，更新是否总是第一个可能的行来进行连接？

create table letter (id_letter bigint, id_group_table bigint, letter char(1));  
create table group_table (id_group_table bigint, id_whatever bigint, champion char(1));
create table whatever (id_whatever bigint);

insert into whatever values (1);
insert into whatever values (2);
insert into whatever values (3);
insert into whatever values (4);
insert into whatever values (5);

insert into group_table values(1, 1, null); -- champion should be B
insert into group_table values(2, 2, null); -- champion should be C
insert into group_table values(3, 3, null); -- champion should be X
insert into group_table values(4, 4, null); -- champion should be C
insert into group_table values(5, 5, null);

insert into letter values(1,1,'A');
insert into letter values(2,1,'B');
insert into letter values(3,1,'B');

insert into letter values(4,2,'B');
insert into letter values(5,2,'C');
insert into letter values(6,2,'C');

insert into letter values(7,3,'X');
insert into letter values(8,3,'X');
insert into letter values(9,3,'Y');

insert into letter values(10,4,'A');
insert into letter values(11,4,'A');
insert into letter values(12,4,'C');

insert into letter values(13,4,'C');
insert into letter values(14,4,'C');
insert into letter values(15,4,'C');

insert into letter values(16,5,'B');
insert into letter values(17,5,'C');
insert into letter values(18,5,'C');

-- update to set champions for everybody

update group_table 
join
  (select  letter,
  id_group_table, 
  count(letter) as occurences
  from letter
  group by id_group_table, letter 
  order by occurences desc
  ) tab on group_table.id_group_table = tab.id_group_table

set champion = tab.letter
where group_table.id_whatever in (1,2,3,4);

http://sqlfiddle.com/#!2/6ea061/1

谢谢！！

Answer 1

documentation 说：

每个匹配的行更新一次，即使它多次匹配条件。

但是，它并没有具体说明它是始终使用的第一个值，所以我不确定您是否可以可靠地依赖它。如果您编写子查询以便它只返回每个id_group_table 的冠军，那会更好。有许多 SO 问题展示了如何编写这样的查询。