TRUNCATE是DDL语句，而DELETE是DML语句。以下是两者的区别:

As TRUNCATE is a DDL (Data definition language) statement it does not require a commit to make the changes permanent. And this is the reason why rows deleted by truncate could not be rollbacked. On the other hand DELETE is a DML (Data manipulation language) statement hence requires explicit commit to make its effect permanent. TRUNCATE always removes all the rows from a table, leaving the table empty and the table structure intact whereas DELETE may remove conditionally if the where clause is used. The rows deleted by TRUNCATE TABLE statement cannot be restored and you can not specify the where clause in the TRUNCATE statement. TRUNCATE statements does not fire triggers as opposed of on delete trigger on DELETE statement

这里有一个非常好的与主题相关的链接。

“Truncate不会记录任何东西”是正确的。我想更进一步:

Truncate不会在事务的上下文中执行。

截断比删除的速度优势应该是明显的。根据你的情况，这种优势从微不足道到巨大不等。

然而，我看到截断无意中破坏了引用完整性，并违反了其他约束。通过在事务外部修改数据而获得的权力，必须与在没有安全网的情况下走钢丝时继承的责任相平衡。

微软sql server的另一个不同之处在于，使用delete，你可以使用output语句来跟踪哪些记录被删除了，例如:

delete from [SomeTable]
output deleted.Id, deleted.Name

你不能用truncate这样做。

TRUNCATE是快的，DELETE是慢的。

尽管TRUNCATE没有问责制。

在这个例子中，Truncate也可以被rollback

begin Tran
delete from  Employee

select * from Employee
Rollback
select * from Employee

它方便的一个重要原因是，当您需要刷新数百万行表中的数据，但又不想重新构建它时。“Delete *”会花费很长时间，而Truncate对性能的影响可以忽略不计。