Nope, but I have seen it before. That's good stuff...

Aha - now I have the context. SQL_VARIANT seems like a reasonable choice now. I'm just a bit confused over the structure...

That is *extreme and soon to be fatal* fragmentation. I used to consider a reboot when we got down below 50MB contiguous.

8MB is not a big block - and...

Indeed. (Don't forget that the article specifically targets SQL Server 7.0 and 2000 though).

Let me put it one last way. There are always alternatives - you may choose...

Garadin (9/11/2009)

---

While the limitations of SQL_Variant are absolutely correct, for his purposes I doubted the existence of any lob types. Timestamp is a possibility, but since he's using date...

I just read it. Um. Trying to be positive here: ok - he...

Ed, please read the link I posted for you.

As far as hardware is concerned, the basic idea is to separate random I/O from sequential I/O. This is the reasoning...

Garadin (9/11/2009)

---

You appear to be using a table variable to do your running totals. You don't want to do this, because you can't establish a clustered index on a...

Garadin (9/11/2009)

---

With that in mind though, consider using sql_variant instead of casting everything to varchars. I say consider because I haven't used sql_variant much myself, so I can't swear...

Yes, because the row data in those columns will end up in the...

Yep. Same code. It's very popular 😀

Two quick comments first up:

NOT EXISTS is faster than (SELECT COUNT(*) FROM table) = 0

Have a look at Erland Sommarskog's article on dynamic searches. It has parallels with what...

Sure: Using Special Data

SYSNAME has been around at least since 6.5 😎

Yeah - I didn't mean to, I just got lucky.

Branching and plan compilation:

http://technet.microsoft.com/en-us/library/cc966425.aspx covers it - the bit beginning "Example: Suppose that a batch B contains an "if" statement..."

http://msdn.microsoft.com/en-us/library/ee343986.aspx is...

Nope 😀