lol owned

good question

Bottom line: always specify whether you want to create a clustered or nonclustered primary key on the [font="Courier New"]add constraint PK_..[/font] statement.

Oleg, thank you very much for the time and explanation, you couldn't have been clearer. i'm gonna play with your code at lunch time :p .

I've learned a lot more...

Thank you Oleg, I am curious about uniquely identifying rows in tables w/o PKs (sorry about the off topic).

I mean, as I understand from your explanation there is some "row...

Hey pal,

what do you mean by volatile rowid from the heap?

Check a couple of posts upstairs, i pointed out the case of a full nulled row, but wasn't sure "why"...

count_big(*) will count every row in table, including rows with null values on every column, probably because it gets the amount of rows out of catalog?

Try this:

declare @t table(i int,...

An administrator changed the possible answers because of the question's subjectivity.

Assuming that:

* the inner select returns a reasonable amount of rows - so it fits in memory

shouldn't the IN statement be better since the inner query executes only once,...