Beware of Mixing Collations - Part 1

  • Greg Larsen

    SSC-Insane

    Points: 20585

    Comments posted to this topic are about the content posted at http://www.sqlservercentral.com/columnists/glarsen/collate_part1.asp

    Gregory A. Larsen, MVP

  • David.Poole

    SSC Guru

    Points: 75043

    I've come across this problem in SQL 6.5.

    Pre-SQL2000 my solution was to have a test machine that I could set up SQL Server 6.5 with whatever character set and sort order I needed to restore other peoples data, then bcp it across onto the production server.

    Pure brute force, but strangely satisfying!

  • Steven.

    SSCrazy Eights

    Points: 9806

    Beware is right, but as long as it is known to the developers etc I have found it isn't too much of a problem.

    I have had to mix collations on a sql 2000 server (Binary sort order server with dbs, plus a single non case sensitive db). Once you remember that tempdb is case sensitive etc it is not too much of a problem.

    This has even had the bonus of helping enforce naming standards since tables in the other DB's and tempdb need to be exact.

    Steven

  • Russel Loski

    SSCertifiable

    Points: 6913

    Can mixed collations affect performance?

    I created a new server and database with a different default collation when I took over our small SQL Server environment. In transferring the objects from the existing server to the new production server, I used the scripts created by EM. For some reason, EM puts collation statements on all character based fields.

    We connect to SQL Server from Access and for several months the applications would crawl. One night I came in and changed all of the collations to use the defaults. That morning the applications ran real fast and it still runs well. I may have done something else as a side consequence, but the only thing I remember is changing the collations.

    Is this a known issue? I don't have the lab facilities to test it.

    Russ

    Russel Loski

    Russel Loski, MCSE Business Intelligence, Data Platform

  • dwestmore

    SSC Veteran

    Points: 258

    I've had nasty problems with the collation settings. The strange thing is I had never explicitly done anything to mess about with collation settings, it seemed to happen spontaneously. I work in Italy, some kind of mess up between English and Italian versions of SQL and/or Windows.

    I did find a script (I think on this site, don't ask me where) for easily changing collation settings on all tables in a DB, if anyone is interested I can send it.

  • tomstraetman

    Grasshopper

    Points: 19

    that would be great, can you send me the script to tom.straetman@pi.be

    thx

  • Greg Larsen

    SSC-Insane

    Points: 20585

    I'd like that script. Can you send it to me at greg.larsen@doh.wa.gov

    Gregory Larsen, DBA

    If you looking for SQL Server Examples check out my website at http://www.geocities.com/sqlserverexamples

    Gregory A. Larsen, MVP

  • Andy Warren

    SSC Guru

    Points: 119655

    Maybe of you guys can either find or repost the script and post the link back here? Others will be interested Im sure.

    Greg, have you looked to see what performance implications of specifying a collation in the join? I use the technique myself when I need a case sensitive join on data that is stored case insensitive.

    Andy

    http://www.sqlservercentral.com/columnists/awarren/

  • Greg Larsen

    SSC-Insane

    Points: 20585

    No I have not checked into the performance issues. I might be interesting to see if there are any issues here.

    If I get the script I will make sure it gets posted.

    Gregory Larsen, DBA

    If you looking for SQL Server Examples check out my website at http://www.geocities.com/sqlserverexamples

    Gregory A. Larsen, MVP

  • Kevin R. Zakoski

    SSC Journeyman

    Points: 94

    We've run into similar problems, but it may be more important to state that it also depends on where and how a #temp table is created in TempDB.

    A 'Create Table #Temp' statement will default the tables collation to that of TempDB's. If you create the #temp table with a SELECT INTO #Temp from a DB with a different collation than TempDB, the #Temp table will have the collation of DB it was run from. The same holds true for regular tables you create outside your DB. When we ran into our problem, it wasn't apparent what exactly was occuring until we removed the '#' from the table name and created static tables which we could look at in more detail to resolve our problem.

    My suggestion is, if you must use different collations on the same server, it's best to adopt a standard for creating #temp tables within your code. If you mix Creates and Select INTO for #Temp tables in your code you will eventually get burnt.

  • Greg Larsen

    SSC-Insane

    Points: 20585

    TEMPDB table do present problem in a mixed collation environment. The TEMPDB takes on the collation of the MSDB database every time the server is restarted. One option might be to set your MSDB collation to the most common collation used, therefore minimizing the issues with collations differences across databases.

    Gregory Larsen, DBA

    If you looking for SQL Server Examples check out my website at http://www.geocities.com/sqlserverexamples

    Gregory A. Larsen, MVP

  • Greg Larsen

    SSC-Insane

    Points: 20585

    TEMPDB table do present problem in a mixed collation environment. The TEMPDB takes on the collation of the MSDB database every time the server is restarted. One option might be to set your MSDB collation to the most common collation used, therefore minimizing the issues with collations differences across databases.

    Gregory Larsen, DBA

    If you looking for SQL Server Examples check out my website at http://www.geocities.com/sqlserverexamples

    Gregory A. Larsen, MVP

Viewing 12 posts - 1 through 12 (of 12 total)

You must be logged in to reply to this topic. Login to reply