Performance 效率更高的是：多个插入件与带有联合件的单个插入件

我在Postgresql中有一个大表（~6M行，41列），如下所示：

id | answer1 | answer2 | answer3 | ... | answer40
1  | xxx     | yyy     | null    | ... | null
2  | xxx     | null    | null    | ... | null
3  | xxx     | null    | zzz     | ... | aaa

请注意，每行中都有许多空列，我只想要那些包含数据的列

我想将其正常化以获得以下结果：

id | answers
1  | xxx
1  | yyy
2  | xxx
3  | xxx
3  | zzz
...
3  | aaa

问题是，几次插入或一次插入和多个联合哪个更有效/快速

选择1

create new_table as 
select id, answer1 from my_table where answer1 is not null
union 
select id, answer2 from my_table where answer2 is not null
union 
select id, answer3 from my_table where answer3 is not null
union ... 

选择2

create new_table as select id, answer1 from my_table where answer1 is not null;
insert into new_table select id, answer2 from my_table where answer2 is not null;
insert into new_table select id, answer3 from my_table where answer3 is not null;
...

---

选项3：有更好的方法吗？

选项2应该更快

将所有语句包装在

begincommit

块中，以节省单个提交的时间

对于更快的选择，请确保要筛选的列（例如，answer1不为null的

）具有索引