How to fix a collation conflict in a SQL Server query?
Adding to the accepted answer, you can used DATABASE_DEFAULT
as encoding.
This allows database to make choice for you and your code becomes more portable.
SELECT MyColumn
FROM
FirstTable a
INNER JOIN SecondTable b
ON a.MyID COLLATE DATABASE_DEFAULT = b.YourID COLLATE DATABASE_DEFAULT
I resolved a similar issue by wrapping the query in another query...
Initial query was working find giving individual columns of output, with some of the columns coming from sub queries with Max or Sum function, and other with "distinct" or case substitutions and such.
I encountered the collation error after attempting to create a single field of output with...
select
rtrim(field1)+','+rtrim(field2)+','+...
The query would execute as I wrote it, but the error would occur after saving the sql and reloading it.
Wound up fixing it with something like...
select z.field1+','+z.field2+','+... as OUTPUT_REC
from (select rtrim(field1), rtrim(field2), ... ) z
Some fields are "max" of a subquery, with a case substitution if null and others are date fields, and some are left joins (might be NULL)...in other words, mixed field types. I believe this is the cause of the issue being caused by OS collation and Database collation being slightly different, but by converting all to trimmed strings before the final select, it sorts it out, all in the SQL.
You can resolve the issue by forcing the collation used in a query to be a particular collation, e.g. SQL_Latin1_General_CP1_CI_AS
or DATABASE_DEFAULT
. For example:
SELECT MyColumn
FROM FirstTable a
INNER JOIN SecondTable b
ON a.MyID COLLATE SQL_Latin1_General_CP1_CI_AS =
b.YourID COLLATE SQL_Latin1_General_CP1_CI_AS
In the above query, a.MyID and b.YourID would be columns with a text-based data type. Using COLLATE
will force the query to ignore the default collation on the database and instead use the provided collation, in this case SQL_Latin1_General_CP1_CI_AS
.
Basically what's going on here is that each database has its own collation which "provides sorting rules, case, and accent sensitivity properties for your data" (from http://technet.microsoft.com/en-us/library/ms143726.aspx) and applies to columns with textual data types, e.g. VARCHAR
, CHAR
, NVARCHAR
, etc. When two databases have differing collations, you cannot compare text columns with an operator like equals (=) without addressing the conflict between the two disparate collations.