Recommendations - Migration for Informix (Magic xpa 2.x)
The migration path offered in this section refers to engines running on the Windows 32-bit platform. The engine can either be a client or server engine.
Data access to Informix will be done via the ODBC capabilities of Magic xpa.
Each engine needs to have an ODBC driver installed.
Some ODBC drivers do not need an Informix client to be installed in order for the driver to function. When using this sort of driver there will not be an additional layer added in between the Magic xpa engine and the database, since it will replace the Informix client.
Data Repository
All columns must not have any native Informix data types set in the Type property (SQL section in Column Properties).
All columns must use the default storage types (Stored As property in Storage section).
Each data source needs to have at least one unique index defined in the Magic xpa Data repository. This is because Magic xpa needs a unique identifier for a single record. Informix supplies this identifier out of the box, but when accessing data via ODBC, you need to define a unique index, where its segments will define the unique identifier.
If there is no known set of segments that can identify a single record, you can adopt the following solution:
-
Add a column to the data source entry and name it ROWID.
-
Define it as a Numeric data type, usually 7 digits.
-
Add a new virtual unique index that includes this column entry only.
-
Define this unique index as the data source's position (Data Source Properties > SQL tab).
Note: The ROWID column exists in Informix tables if they are non-fragmented. It does not exist in fragmented tables nor in certain views.
It is recommended by IBM to use primary keys to access records rather than ROWIDs.
Program Repository
The application might have Direct SQL statements that will have to be revised once using ODBC. Examples for this might be SELECT statements or Stored Procedures.
The volume of the changes needed in this case is unpredictable since it completely depends on the application specifics.
For data sources that use the ROWID as a position, you need to add the ROWID to the data view of all tasks that actually insert records into this data source (either as a main source or linked data source). All other tasks can be left as is.
