Performance Issue with Internal Tables

[Former Member]

Hi,

I'm confronted with the following problem. The source table containing datafield Product-Count (/B99/S_BWPRDTANZ) is incorrectly filled i.e. if there are two rows containing the same Product (/B99/S_BWPRDT) and Product-Group (/B99/S_BWPRDGRP) both rows have value 1 entered for Product-Count. It should actually be 1 in the first row and 0 in the second row.

I am basically sorting the source and filling Table IT_XX_TAB1 with the first record with CRM-Object-ID (CRM_OBJ_ID) as the key. The rest of the records are filled into Table IT_XX_TAB2. Both tables are compared for a particular CRM-Object-ID and the result is updated accordingly.

The code below works fine for the few sample records I took while testing my logic. But for the entire set of records it ran for around 11 hours until I got the result. Any suggestions as to how I could improvise on the code and performance of the routine? It is an APD routine with a total of 9972295 Records.

  DATA: ls_source TYPE y_source_fields,

        ls_target TYPE y_target_fields.

  DATA: INDEX TYPE i VALUE 0,

            FLAG TYPE i.

  DATA: IT_XX_TAB1 TYPE STANDARD TABLE OF y_target_fields

        WITH NON-UNIQUE KEY CRM_OBJ_ID,

        IT_XX_TAB2 TYPE SORTED TABLE OF y_target_fields

        WITH NON-UNIQUE KEY CRM_OBJ_ID.

  DATA: WA_XX_TAB1 TYPE y_target_fields.

  SORT it_source DESCENDING BY CRM_OBJ_ID

                               /B99/S_BWAZLVL1

                               /B99/S_BWAZLVL2

                               /B99/S_BWAZLVL3

                               /B99/S_BWAZLVL4.

  CLEAR FLAG.

  CLEAR INDEX.

  LOOP AT it_source INTO ls_source.

    IF ls_target-CRM_OBJ_ID = ls_source-CRM_OBJ_ID.

      INDEX = INDEX + 1.

      FLAG = INDEX.

    ELSE.

      INDEX = 1.

      FLAG = INDEX.

    ENDIF.

    MOVE-CORRESPONDING ls_source TO ls_target.

    IF FLAG EQ 1.

      APPEND ls_target TO IT_XX_TAB1.

    ELSE.

      INSERT ls_target INTO TABLE IT_XX_TAB2.

    ENDIF.

  ENDLOOP.

  et_target = IT_XX_TAB1.

  LOOP AT IT_XX_TAB2 INTO ls_target.

    READ TABLE IT_XX_TAB1 INTO WA_XX_TAB1

    WITH TABLE KEY CRM_OBJ_ID = ls_target-CRM_OBJ_ID.

    IF sy-subrc = 0.

      IF ( ls_target-/B99/S_BWPRDGRP = WA_XX_TAB1-/B99/S_BWPRDGRP

         AND ls_target-/B99/S_BWPRDT = WA_XX_TAB1-/B99/S_BWPRDT ).

        FLAG = '0'.

      ELSE.

        FLAG = '1'.

      ENDIF.

      IF FLAG = 0.

        ls_target-/B99/S_BWPRDTANZ = '0'.

      ENDIF.

      APPEND ls_target TO et_target.

      CLEAR ls_target.

    ENDIF.

  ENDLOOP.

Thanks,

SD

[former_member184455]

Hi SD,

you defined IT_XX_TAB1 as STANDARD table and IT_XX_TAB2 as SORTED table, but you do a LOOP w/o WHERE clause on IT_XX_TAB2 (no optimized access needed) and a READ TABLE ... WITH TABLE KEY on IT_XX_TAB1 (optimized access desperately needed).

So the definition should be the other way round.

Best Regards,

Randolf

[Former Member]

Hi,

are you sure that this is where the performance problem lies?

I think looping on 10 million records shouldn't take that long.

Maybe something is wrong with your memory sizing.

You can use Get timestamp statements to find out exactly how long a statement or loop takes.

Kind regards, Rob Dielemans