How many times will internal table get filled?

Question

Dear BW / ABAP gurus:
I am observing a routine in a transformation in 7.0 data flow.
The internal table is created as follows in the global declaration in the Start Routine:
TYPES:  BEGIN OF ST_PROD,
          /BIC/AZPROD TYPE /BIC/OIAZPROD,
          /BIC/AZCOST TYPE /BIC/OIAZCOST,
        END OF ST_PROD.
DATA: IT_PROD TYPE TABLE OF ST_PROD,
      WA_PROD TYPE ST_PROD.
The programmer is fetching data for product cost which is not available in the source. This code is written in the Start Routine:
SELECT /BIC/AZPROD /BIC/AZCOST
  FROM /BIC/PAZPROD
  INTO CORRESPONDING FIELDS OF TABLE IT_PROD
  WHERE OBJVERS = 'A'.
My questions:
(1) Is it a standard practice to create the internal table in the global declaration?
(2) Is it a standard practice to fill the internal table with data in the Start Routine? Is it possible to fill the internal table with data in the global declaration? Or is the global declaration only reserved for declaring variables and creating internal tables?
(3) What is the use of 2nd part global?
(4) Let's say there are 100, 000 records at the source and the DTP package size is 50,000. Then the Start Routine will get executed 2 times. So the code in the Start Routine will hit the database 2 times to fill the internal table. Is this correct?
Note to mods: Please do not delete the thread. Move to the beginner section if you think this is basic. But please do not delete. It take a long time to type out the questions.

Answer 1

hi Saurav,
(1) Is it a standard practice to create the internal table in the global declaration?
1. No ,internal table should be created in global part only if they are getting used in field level or end routine .Else they must be in local part and should get refreshed at the end  so that performance not get affected.
(2) Is it a standard practice to fill the internal table with data in the Start Routine? Is it possible to fill the internal table with data in the global declaration? Or is the global declaration only reserved for declaring variables and creating internal tables?
1.No if only you want to use the data in field level routine and start routine ,you will fill table in start routine otherwise no .
2.first global part is for data declaration so no in first global part select cannot come .
3.Second global part can have select but you cannot write any statement using source package or result package here as they are private object of transformation class and not recognized at this area.However simple select on any table other than these will be fine .
(3) What is the use of 2nd part global?
1.In second global part you can declare internal table ,structures that are not of type source package or result package but independent and can use them .
2.If you will declare these table in first global part they will be private object and you cannot write select using them in second global part .
3.Second global part is also declaration but outside the class so these objects are global across transformation with more flexibility .
4.If you will declare any data here using the structure of source /result package you will get error .This is a data declaration part comes between transformation definition and implementation part .
(4) Let's say there are 100, 000 records at the source and the DTP package size is 50,000. Then the Start Routine will get executed 2 times. So the code in the Start Routine will hit the database 2 times to fill the internal table. Is this correct?
yes the source and result package are actually the DTP packages only so code will get executed same no of time as you have number of packages in DTP extraction .
Hope this will be helpful .
Regards,
Jaya Tiwari