Comments on: Querying MS SQL Server G/UUIDs from ColdFusion

By: Judah McAuley

Judah McAuley — Fri, 06 Mar 2009 22:55:18 +0000

Jochem, just an fyi, but there is a cfsqlparam datatype specified for uniqueidentifiers. It is CF_SQL_IDSTAMP I would not have known but I was just running Illudium against a table with a uniqueidentifier and noticed a weird cfsqlparam that I hadn’t seen before.

http://livedocs.adobe.com/coldfusion/7/htmldocs/wwhelp/wwhimpl/common/html/wwhelp.htm?context=ColdFusion_Documentation&file=00000317.htm

By: Tjarko

Tjarko — Thu, 18 Dec 2008 09:15:39 +0000

@Jochem What I mean is that when i looked at some sites we are running I had several statements “tuned” with cfqueryparam and after those changes the site went from tens of milliseconds to hundreds of milliseconds. Changing it back after several days returned everything to tens of milliseconds. These sites have about 10 to 15 thousand visitors every day.. and that makes a difference in performance.. hence “everything a lot faster” The only difference for this was the use of cfqueryparam.

By: Orange is my favorite color » Blog Archive » Converting from UUIDs to GUIDs

Tue, 16 Dec 2008 18:12:41 +0000

[...] been reading through a handful of Jochem’s posts about Adobe’s (incompatible with the standard 35-character) [...]

By: Jochem

Jochem — Tue, 16 Dec 2008 09:40:27 +0000

@Tjarko: I think you not extrapolate a benchmark I wrote to prove a specific point to “everything” and “a lot faster”.

By: Jochem

Jochem — Tue, 16 Dec 2008 09:38:12 +0000

Numeric or integer? I expect numeric to perform about the same as a uniqueidentifier because, although it is smaller, it is a ‘complex’ type: it has some internal representation that needs to be decoded before any operation can be performed on it.

An integer will be a bit faster because it is smaller. Not that much though, when I went from a varchar to a uniqueidentifier I already got most of the gains becuase that conversion gets rid of all the extra internal operations that are involved with manipulating strings (locale, case sensitivity etc.). So all the gains going from uniqueidentifier to integer stem from the difference in storage size, and that matters less.

But this benchmark is designed for a specific situation (real data from a production system that already has UUIDs that can not be shortened to integers) and to prove a specific point (I benchmark tables with predominantly UUID columns), so it would not be very suited to benchmark anything else.

By: Tjarko

Tjarko — Tue, 16 Dec 2008 07:19:54 +0000

Really curious about your next findings, i always had the sneaking suspicion that without cfqueryparam everything went a lot faster, and apparently this is true.

By: Zac Spitzer

Zac Spitzer — Tue, 16 Dec 2008 02:04:54 +0000

I’d be interested to see the stats where you just use a plain old (of course not replicate friendly) numeric key