nchar(36) value causing conversion failure when attempting to update to a uniqueidentifier column
Posted on 2008-10-08
I have a database with over 40 tables. The PK's and FK's were originally defined as nchar(35) to be compatible with the GUID format of ColdFusion. The middleware for the project has now been changed to .NET and the data architect wishes the nchar(35)'s to be converted over to uniqueidentifier columns in SQL Server. These tables are currently filled with data.
I have been working on this project and am nearly complete with the TSQL Script to convert the entire database, but have run into a very strange situation.
An nchar(35) ColdFusion GUID looks like this:
A SQL Server UNIQUEIDENTIFIER looks like this:
Note the only difference is the hyhen at position 19. Additionally, the uniqueidentifier uses lower case, not upper case. The length of the former is 35, the length of the latter is 36.
To make a long story short, in order to affect the conversion, after dropping all the FK and PK constraints, I follow these steps:
* modify the column containing the GUID from nchar(35) to nchar(36)
* update each value to insert the additional hyphen at position 19
* rename the column to nameID_OLD
* add a new column to the table named ID of type uniqueidentifier
* run an update command to update the ID column with the values in the nameID_OLD column.
My problem is that I have come across a value that is causing SQL Server to report a "Conversion failed when converting from a character string to uniqueidentifier." error as I am running the following command:
set manufacturerID = lower(manufacturerID_OLD)
--manufacturerID is of datatype uniqueidentier
--manufacturerID_OLD is of datatype nchar(36)
Nearly all values will convert implicitly to the uniqueidentifier perfectly. However this value: df0351fa-34f8-49b0-b630-f1294324398t will reliably throw the conversion failure error.
Is there something about this value that causes SQL Server to see this as an invalid uniqueidentifier? What is the spec behind a uniqueidentifier such that I might write a regular expression to locate incorrectly formatted uniqueidentifiers?