De los Tilos sabiduría

bases += patrones += relaciones += datos

Build history models with Oracle SQL Developer Modeler 4.0, intro

without comments

Oracle SQL Developer Modeler (OSDM) is a nice free data modeling tool with a lot of nice features. I’m going to use a set of these features to create historic versions of a sample model. Hans Hultgren and Martijn Evers have made a classification of Ensemble Modeling Forms or Styles of Data Vault modeling historical data warehouse modeling styles. The classifications I will use:

  • – Classical (a.k.a. Dimensional) Data Vault of Dan Linstedt
  • – Anchored Data Vault (a.k.a.) Anchor Vault
    • – the strong version , with no end dated links
      • used in the open source dwh automation tool Quipu and explained by my college Lulzim
    • – the weak version, with end dated links
    • elementary Anchor Vault, all attributes are split in separate tables except the business key
    • Anchor Modeling, all attributes are split in separate tables including the business key
      • Focal point modeling, externalizes the business key but groups attributes

    I’m going to build two versions of a weak Anchor Vault version:

    • a Head Version model, this is the maximum grouped version of an Anchor Vault
    • a ‘regular’ Anchor Vault, this is a less grouped version where we split the foreign keys and the attributes

    The last one will be an Anchor Model, the most split version of them all.

    The general approach used for all the three models in OSDM is:

    1. reverse engineer a (source) model in OSDM into a Relational model
    2. then forward engineer the relational model a to logical model
    3. then in the logical model ‘surrogate’ the model

    This is the starting point for all of the models and is equal for all of the models. The next steps are used for all of the three variants, but are slightly different for each model type:

    1. then create a new relational model with one table that contains the ‘default columns’
    2. then forward engineer the surrogated logical model to the new relational model and use the table with the ‘default columns’ as template
    3. split the tables using the split table wizard to get the correct tables
    4. extend the primary key of the ‘version/history’ tables

    The nice thing of this approach in OSDM is that at the end we have a mapping between the ‘source model’ and the ‘historical model’ via the logical model.
    Let’s get started with the first three generic steps. First we have a DDL of the source model:

    CREATE TABLE Department
    ( DepName VARCHAR (255) NOT NULL
    , Budget  DECIMAL (12,2)
    ALTER TABLE Department ADD CONSTRAINT Department_PK PRIMARY KEY ( DepName )
    CREATE TABLE Employee
    ( EmpName      VARCHAR (255) NOT NULL
    , Gender       CHAR (1)
    , DepName      VARCHAR (255) NOT NULL
    , Job          VARCHAR (255)
    , HoursPerWeek SMALLINT
    , Manager      VARCHAR (255)
    ALTER TABLE Employee
    PRIMARY KEY ( EmpName )
    CREATE TABLE Interest
    ( EmpName  VARCHAR (255) NOT NULL
    , ProdName VARCHAR (255) NOT NULL
    , Degree   SMALLINT
    ALTER TABLE Interest
    PRIMARY KEY ( EmpName, ProdName )
    CREATE TABLE Product
    ( ProdName VARCHAR (255) NOT NULL
    , Price    DECIMAL (12,2)
    ALTER TABLE Product
    PRIMARY KEY ( ProdName )
    ALTER TABLE Employee
    ADD CONSTRAINT Employee_Department_FK 
    FOREIGN KEY (DepName ) REFERENCES Department ( DepName ) ;
    ALTER TABLE Interest
    ADD CONSTRAINT Interest_Employee_FK FOREIGN KEY ( EmpName
    ) REFERENCES Employee ( EmpName )
    ALTER TABLE Interest
    ADD CONSTRAINT Interest_Product_FK
    FOREIGN KEY ( ProdName ) REFERENCES Product ( ProdName )
    ALTER TABLE Employee
    FOREIGN KEY ( Manager ) REFERENCES Employee ( EmpName )

    1. Import the DDL file into a Relational model. I used the DB2/UDB 7.1 setting to import this DDL correctly. The result is a nice model:

    Ok, now we have the model in OSDM.
    2. We can forward engineer it to a logical model.

    3. The last generic step is to surrogate the model. I made a nice JavaScript you can use to automate this part. You can add this as a custom transformation script:

    In the screenshot above the ‘Mozilla Rhino’ engine is not there but the ‘Rhino’ engine is there on my Ubuntu machine. Somehow the same javascript engine it is reported different. The logical model know looks like this:

    The model in this state is preserved in this GitHub repository as master.
    All the next versions will be saved as a branch of the master model. That’s it for now. The next posts will be:

    Written by delostilos

    April 8th, 2014 at 9:27 am

    Leave a Reply