Question

I know that within a trigger – at least for SQL Server – one should never assume that the inserted table has just one row, which means SQL like this in a trigger is usually bad:

select @UserID = ID from inserted

But out of curiosity, can a set of INSERT statements ever result in an inserted table of more than one row? I know it’s easy enough with an UPDATE, but from my tests I can’t simulate a similar result for INSERTs. I’ve tried sending across sets of inserts before sending the batch terminator, e.g:

insert into TriggerTest (col2) select 'a'
insert into TriggerTest (col2) select 'b'
insert into TriggerTest (col2) select 'c'
go

And also wrapping them in transactions:

begin tran
insert into TriggerTest (col2) select 'a'
insert into TriggerTest (col2) select 'b'
insert into TriggerTest (col2) select 'c'
commit

But it will always result in the trigger fired 3 times with an inserted table of 1 row, and never one time with an inserted table of 3 rows.

That completely makes sense to me (they are 3 separate statements after all), and I don’t need to actually do it, I’m just wondering if INSERTS alone can ever behave differently to this.

Edit: this is a dumb question: of course it can, when inserting a result set!

insert into TriggerTest (col2) select 'a' union select 'b'

… or any other sort of set.

Forgive me, it is almost 3AM here. I’ll leave this question here for people who should know better anyway.

Answer 1

try

insert into TriggerTest (col2) 
select 'a'
union 
select 'b'
union
select 'c'