Custom databases (oracle tables versus abap 'z' tables)

Question

We are relatively new to SAP PI.  Many of our developers come from a EAI background rather than an abap background.  How to do you recommend creating customized tables - directly create an oracle database and schema and administer outside of the sap landscape - retrieving data from java or by creating the oracle tables from within the abap stack?
Are there any performance considerations in one approach over another?
Given the fact that SAP long-term plans on moving away from the abap stack - is that a consideration?

Answer 1

Hi,
Imho, you need to get (meaning, extract into separate fields) the different supplier types from Table1 first. Your key for Table1 is the vendor no, which is also the key in Table2 (or the key for Table2 is Vendor no & Type).
For better performance, better select multiple/all required entries from Table1 instead of doing a select endselect.
Depending on the format of the vendortypes in Table1, put them in a new itab (for our purpose named Table1New where vendor no & type are the only 2 fields. For example, if the type length is fixed to 2 chars, or divided by space,... use your coding accordingly.
Next step is to select all vendor no's in Table2 which you have selected in Table1. If in Table2, the vendor no is the only key (and the all vendor types are filled in a single record), then loop check the vendor types from Table1New against the types in Table2.
If the key of Table2 is vendor no & vendor type, then do a read table for the key.
The logic in pseudo-code:
Select from Table1 into table. If you'd like to limit the selection size, add package size statement.
     extract the vendor types in to itab Table1New.
     Select the vendor & types from Table2 by using the for all entries option (better performance).
     loop at Table1New
          check in Table2:
               if the unique key is vendor no: check all fields for the vendor type from Table1New
               if the unique key combo is vendor no & type: check by using a read table.
          If not found => add entry to Table2
     endloop.
endselect Table1 (when using package size)
I guess the most difficult step is to extract the types from Table1 into separate fields, all the rest seems straight forward. Please keep in mind the itab type definitions for a better performance.
Good luck!
Best regards,
Zhou