I bring up the lack of randomness as it works in the favor of delimitedsplit8k, as the optimizer can basically get to reuse its row plans. (Running the test script...

Eirikur Eiriksson (5/22/2015)

Just ran a quick...

Sure Jeff, I've added the generator and the full test script into my gist.

Here is a direct link to the script:

https://gist.github.com/mburbea/5e142f846d0141c714a1#file-testscript-sql

This was run against a Sql server 2014 instance....

Most of the methods described are worse than delimitedsplit8k. The CLR splitter is decent. And TVPs are unusual and unfortunately, unwieldy to use without storeprocedures and non-normal client code. They...

I discovered something interesting today using the LAG and LEAD function. You can use them in TVFs to create cached versions of ctes and intermediary results, the trick is to...

Solomon Rutzky (9/9/2014)

Regarding the two items you mentioned, Tally function and splitter, both are available for free in the SQL#[/url] library, along with quite a few others. There are several...

To be fair, Microsoft did give us master data services which has a CLR udf splitter (an awful one) that still requires enabling CLR on so most DBAs just say...

It seems that Paul is right, which explains why it appears that the memory optimized tally tables splitter beat CLR functions at first blush. (This isn't true sadly). Adding the...

If you look at my github I included an update test script which should make it easier to test your functions. Simply add them to the function array on top...

For those looking for the fastest way to split 8k CSVs in SQL Server, I present hybrid.

Hybrid performs about TWICE as fast as Paul White's Split CLR function. And about...

Paul's CLR split method is quite excellent and so simple. However, in this trial part of its pain points is that it takes an nvarchar(max) and so every call has...

Solomon Rutzky (5/20/2014)

Regarding recycling the App Domain, here is a little known trick that I stumbled upon while testing something a couple of years ago: ALTER an Assembly in that...

One thing I discovered is that Jeff's test harness is a bit biased at least when it comes to the CLR functions.

I was trying to write a faster CLR...

Thank you Paul for figuring out why this benchmark was suddenly flying after upgrading from 2008 R2 to 2014. It's still nice and fast but a lot more reasonable.

Out of...

Usman's reformation of the query to remove the union from "cteStart" and add it to the final result as has a nice appreciable gain in performance. It now seems to...