@sergiy: No, if these are inline UDFs they should be expanded in place, similar to a C macro and shouldn't incur a RBAR-like performance hit.

Regards,

Jacob

I'd create 4 seperate inline UDFs, one for each DataField column. Each function would take in two parameters (RecordType & DataToBeParsed) and return a varchar(10) value, and would encapsulate...

The same is true (though to a lesser extent) with any of the algorithms supported by HashBytes(). There is no guarantee that two of your rows that have the...

D'OH! Jan already posted while I composed my reply - ignore this....

Check out either CHECKSUM() or HashBytes() in BOL. CHECKSUM's been around for ages and computes a CRC (CRC32...

Ouch! Yep, that does the trick. I wonder if there will ever be a situation where the merge join would be suboptimal. Thinking about it, probably only if there...

I can verify your results: changing to the dynamic SQL version runs at 1.2 sec at @PageNum = 8000 on my test data/server.

Now to speculate on why...

I can't compare the...

@ Matt:

Yep - ran the same test as above in my reply to Jeff, same data, same server except using your test code. Avg...

@ Jeff: Is this

a real-world figure? I put...

You can manually merge cells in RS (just like Excel), but not dynamically afaik. Probably not what you want...?

Regards,

Jacob

@ Jeff:

Just need RowNum on the middle select don't you (as opposed to *)?:

...

SELECT t.*

FROM dbo.Test t,

(

...

OK - first of all you can't do this:

...WHERE a.ParentPartCode IN (@PART)

unless you manually expand @PART into it's value using dynamic SQL. Have a read of Erland Sommarskog's excellent...

A CLR-based approach to this may be best - build a work-breaker function that takes your Source A and Source B strings as input. I'm making an assumption that...

Ummm... add a second report parameter?

:ermm:

Regards,

Jacob

100% agree with ironpants. You should start with RI enforced always, and only ever consider removing it in special cases (usually for performance issues that can't be solved any other...

What Steve Jones said, except that I wouldn't entirely rule out RAID5 - it's suitable for a workload heavily biased towards reads with little/no write activity and lower recovery SLAs,...