将SQL Server数据库中的所有varchar列更改为nvarchar

Question

I am trying to update all varchar columns in a database to nvarchar without losing the data.

I have the following query:

select * 
from INFORMATION_SCHEMA.COLUMNS 
where DATA_TYPE = 'varchar'

How do I modify this to update the columns it finds and have that reflected throughout the database?

Answer 1

I would use the SQL Server specific sys catalog views, and what I'd do is just simply create the T-SQL statements it takes to change those data types - something like this:

SELECT
    t.name, s.name, c.name,
    'ALTER TABLE ' + s.name + '.' + t.Name + ' ALTER COLUMN ' + c.name + ' NVARCHAR(' + 
        CASE 
            WHEN c.max_length = -1 THEN 'MAX'
            WHEN c.max_length > 4000 THEN 'MAX'
            ELSE CAST(c.max_length AS VARCHAR(10)) + ');'
        END
FROM
    sys.tables t
INNER JOIN 
    sys.schemas s ON t.schema_id = s.schema_id
INNER JOIN 
    sys.columns c ON c.object_id = t.object_id
INNER JOIN 
    sys.types ty ON c.system_type_id = ty.system_type_id
WHERE
    ty.name = 'VARCHAR'
    AND c.is_computed = 0

Run on the AdventureWorks database, this will output something like:

ALTER TABLE Sales.SalesOrderHeader ALTER COLUMN CreditCardApprovalCode NVARCHAR(15);
ALTER TABLE Person.Password ALTER COLUMN PasswordHash NVARCHAR(128);
ALTER TABLE Person.Password ALTER COLUMN PasswordSalt NVARCHAR(10);

which you can then copy from the output grid into a new query window and execute.

Note: running your original script against AdventureWorks showed me that in that case, one of the columns is computed and can therefore not be modified by simply using a ALTER TABLE ... ALTER COLUMN .... statement - that's why I'm excluding computed columns in my script shown here.

If you really insist, you could also turns this SQL statement of mine into the "heart" of a cursor and then iterate over the T-SQL statement to modify the table, and execute those using EXEC(...) or EXEC sp_executesql