Creating User-Defined Data Types - SQL School Video

  • I've been using UDTs for years for those very reasons of consistency & self-documentation you propounded in your video. Updating the UDT has always been its Achilles heel.

    Currently

    You suggest in the video that UDTs should be assigned retrospectively once the database design has settled down - which almost contradicts their reason to exist! Nevertheless, structures DO change and we must synchronise our types to the new design. Here's a handy 2005 script showing all dependencies for your UDTs. At least you can see which objects you need to change or drop before updating your type. This script could be wrapped into a parameterised SP for refined filtering etc.

    use [YourDB]

    Select t.name UDT,

    o.name Object,

    o.type_desc ObjectType,

    c.name AttributeName,

    c.type_desc AttributeType

    From (

    SELECT object_id, name, user_type_id, 'SQL_COLUMN' AS type_desc

    FROM sys.columns

    UNION ALL

    SELECT object_id, name, user_type_id, 'SQL_PROCEDURE_PARAMETER'

    FROM sys.parameters

    ) c

    JOIN sys.objects o ON o.object_id = c.object_id

    JOIN sys.types t ON t.user_type_id = c.user_type_id

    Where t.schema_id = 1-- dbo (see sys.schemas for all other object owners & change as reqd)

    order by

    UDT,

    Object

    and here's one for SQL Server 2000.

    use [YourDB]

    Select t.name UDT,

    o.name Object,

    t.xusertype ObjectType,

    c.name AttributeName,

    case o.xtype

    when 'P'

    then 'SQL_PROCEDURE_PARAMETER'

    else 'SQL_COLUMN'

    end AttributeType

    From syscolumns c

    JOIN sysobjects o ON o.id = c.id

    JOIN systypes t ON t.usertype = c.xusertype

    Where t.uid = 1-- dbo (see sysusers for other users' IDs)

    and o.xtype in ('P','U')

    order by

    UDT,

    Object

    Another thing about UDTs: you can't use them in the [font="Courier New"]Convert()[/font] or [font="Courier New"]cast()[/font] functions! You get "Type xxx is not a defined system type". When are MS going to get that right? (Apologies if this exists in 2008.)

    Proposal

    But life shouldn't be that hard, and I would propose the following solution for a near-future SQL Server upgrade.

    Eliminate UDTs!

    In almost all database scripting scenarios, every variable is derived from, or has a relationship with, a table-based field. Why not use that field's type when declaring any variable or SP parameter etc? So, for example, when creating a variable relating to [font="Courier New"]tblClient.EMailAddress[/font], it would be defined as:

    [font="Courier New"]declare @EMailAddress tblClient.EMailAddress%Type[/font]

    The upshot of this extension to the syntax is that all variables' types would be dynamic and would depend on the CURRENT type of the underlying source field at execution time. No need to update anything except the table structure! In the purest form, no type except for those defining table fields would use a system type. Oracle have been doing it for years and it works a treat. From a user/programmer perspective, the change would be painless, rock solid and even more self-documenting than UDTs now. Changing the underlying SQL Server engine to use this proposed syntax may not be be so trivial...

  • I think you are dead on regarding the cumbersome nature of the udt's and about Oracle's typcasting of variables. Wouldn't it be nice to have a stored procedure where you feed in the user defined datatype to be modified as one argument, and the revised definition as the second argument, where upon the sp produces a script that finds all instances and generates the necessary conversion code? (or an SSIS package to do this)

    After being diligent over the years to define domain types in erwin and its ilk, this seems like it could be really handy if it weren't so cumbersome.

    SCott

Viewing 3 posts - 1 through 3 (of 3 total)

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