DB Design Help

  • I am creating a change management system. Part of my design has a change request table and a contact table. Each change request has a technicianid, a managementid, a directorid, and a maincontactid. To find out all the info on the people for a change request, I have to join to the contact table four times.

    From the contact table, I have a department table. So I have to join another four times to find the department of all the contacts for a single change.

    Does this sound like a proper database design? Including the domain tables, I am up to 11 joins to find out the data I want on a particular change request.

  • No, that's probably not a great way to go about it. Instead have a many-to-many join table between the two other tables. So you'll have a ChangeRequest table and a Contact table. Then you create a ChangeRequestContact table that has they key from each of the other two in it. This table should also have a link to a lookup table ContactType (or something, you name these things for your business). That way you can define how a given contact is associated with the change request. This will allow for you to add more and more contact types without having to redesign the system. I'd do the same thing for the department so that more than one department can be easily associated with a given request in varying roles. It makes for somewhat more complicated coding, but it's a lot more flexible and allows for change without having to re-structure your db and rewrite your code.

    "The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood"
    - Theodore Roosevelt

    Author of:
    SQL Server Execution Plans
    SQL Server Query Performance Tuning

  • Grant Fritchey (2/27/2009)


    No, that's probably not a great way to go about it. Instead have a many-to-many join table between the two other tables. So you'll have a ChangeRequest table and a Contact table. Then you create a ChangeRequestContact table that has they key from each of the other two in it. This table should also have a link to a lookup table ContactType (or something, you name these things for your business). That way you can define how a given contact is associated with the change request. This will allow for you to add more and more contact types without having to redesign the system. I'd do the same thing for the department so that more than one department can be easily associated with a given request in varying roles. It makes for somewhat more complicated coding, but it's a lot more flexible and allows for change without having to re-structure your db and rewrite your code.

    1) For the old database design, writes are fast and reads would be a little slower since they have to do so many joins. I have to build a report each month that shows the changes for that month. I figure 100 rows a month selected. Wouldn't read performance on only 100 rows be acceptable even if the database starts holding tens of thousands of rows?

    2) I realize the design you suggest is much more flexible. But wouldn't I still have to do the same amount of joins to select the data? Below would be the insert statement. (assuming I don't have the ContactType lookup table)

    CREATE PROCEDURE InsertChangeRequest_sp

    (@TargetSystem varchar(40), @Description varchar(1000),

    @TechnicianID int, @ManagementID int,

    @DirectorID int, @MainContactID int)

    AS

    BEGIN TRAN

    INSERT INTO ChangeRequest

    (TargetSystem, Description)

    VALUES (@TargetSystem, @Description)

    DECLARE @ChangeRequestID int

    SELECT @ChangeRequestID = Scope_Identity()

    INSERT INTO ChangeRequestContact

    (ChangeRequestID, ContactID, ContactType)

    VALUES

    (@ChangeRequestID, @TechnicianID, 'Tech')

    INSERT INTO ChangeRequestContact

    VALUES

    (@ChangeRequestID, @ManagerID, 'Manager')

    INSERT INTO ChangeRequestContact

    VALUES

    (@ChangeRequestID, @DirectorID, 'Director')

    INSERT INTO ChangeRequestContact

    VALUES

    (@ChangeRequestID, @MainContact, 'MainContact')

    END TRAN

    -------------

    Then the select statement would be

    SELECT CR.TargetSystem, CR.Description,

    (CT.FirstName + CT.LastName) as TechnicianName,

    (CM.FirstName + CM.LastName) as ManagerName,

    (CD.FirstName + CD.LastName) as DirectorName,

    (CMC.FirstName + CMC.LastName) as MainContactName

    FROM ChangeRequest CR

    JOIN ChangeRequestContact CRC ON CR.ChangeRequestID = CRC.ChangeRequestID

    JOIN Contact CT ON CRC.ContactID = CT.ContactID

    JOIN Contact CM ON CRC.ContactID = CM.ContactID

    JOIN Contact CD ON CRC.ContactID = CD.ContactID

    JOIN Contact CMC ON CRC.ContactID = CMC.ContactID

    --plus four more joins for department

    WHERE ScheduleDate is this month

    Does this sound right?

  • Yeah, you'll have a few joins, it's still not that many and the indexing should be very simple. The one thing I'd suggest, instead of trying to pivot the data into columns, return it as rows, you'll have fewer joins and simpler queries that way. You can change the layout on the client side so that it's pivoted.

    "The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood"
    - Theodore Roosevelt

    Author of:
    SQL Server Execution Plans
    SQL Server Query Performance Tuning

  • The ChangeRequestContact table is the best way to handle this type of data. As an added daat integrity measure i would place a constraint on the table only allowing one contact per role type per change request.

    Stephen Marais
    Integration Architect
    Digiata Technologies
    www.digiata.com

  • Grant Fritchey (3/6/2009)


    The one thing I'd suggest, instead of trying to pivot the data into columns, return it as rows, you'll have fewer joins and simpler queries that way. You can change the layout on the client side so that it's pivoted.

    Can you explain this a little more? I am not following how I could pivot into columns instead of rows. What would my query look like?

  • The data would be something like

    Name, Type

    Joe Plumber

    Tim Project Manager

    Tom Developer

    Etc.

    Instead of what you're trying which is:

    PlumberName, Project Manager Name, DeveloperName

    Joe Tom Tim

    "The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood"
    - Theodore Roosevelt

    Author of:
    SQL Server Execution Plans
    SQL Server Query Performance Tuning

Viewing 7 posts - 1 through 6 (of 6 total)

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