Skip to end of metadata
Go to start of metadata

Author: Karthik C Sunil
Submitted: 18-Aug-2007
Related Links:


This page will explain how to code very efficiently for nested un-related tables. The article describes how to improve the performance of processing huge data amounts.

Problem Description

The most common performance problem that occurs in ABAP programs is because of huge number of records in the internal tables. The problem complexifies if a program has huge nested internal tables. How much ever efficient data select routines are, data processing routines would be contributing significantly for the bad performance. When analysed it would be revealed that the where condition that is used in inner loops expend a significant amount of processing time. The idea is to avoid where conditions in the inner loops by maintaining the loop indexes manually.

Conventional Code for nested loops

Error rendering macro 'code': Invalid value specified for parameter 'com.atlassian.confluence.ext.code.render.InvalidValueException'
loop at lt_vbpa into wa_vbpa.
  loop at lt_kna1 into wa_kna1 where kunnr = wa_vbpa-kunnr.

****** Your Actual logic within inner loop ******


Code sample: Parallel Cursor method

Error rendering macro 'code': Invalid value specified for parameter 'com.atlassian.confluence.ext.code.render.InvalidValueException'
sort: lt_vbpa by kunnr,  "Sorting by key is very important
      lt_kna1 by kunnr.  "Same key which is used for where condition is used here
loop at lt_vbpa into wa_vbpa.
  read lt_kna1 into wa_kna1     " This sets the sy-tabix
       with key kunnr = wa_vbpa-kunnr
       binary search.
  if sy-subrc = 0.              "Does not enter the inner loop
    v_kna1_index = sy-tabix.
    loop at lt_kna1 into wa_kna1 from v_kna1_index. "Avoiding Where clause
      if wa_kna1-kunnr <> wa_vbpa-kunnr.  "This checks whether to exit out of loop

****** Your Actual logic within inner loop ******

   endloop. "KNA1 Loop
endloop.  " VBPA Loop

Statical Analysis

Nested loop for BSEG and BKPF internal tables were analysed for Conventional Method and Parallel Cursor methods. Following gragh explains the observations.

Observation: One can observe that as the data increases, the time taken for the nested loop increases drastically, at the same time, the Parallel cursor method did not suffer any considerable time impact.

Verdict: Use the parallel cursor method whenever there is a need to process data in a nested loop.

1 Comment

  1. Former Member

    1) I am not sure if the sort on table lt_vbpa is required at all.

    2) SE30 Documentation talks about O (n1+n2) runtime for a parallel cursor algorithm.

    The documentation extract:

    Nested loops Documentation
    If ITAB1 has n1 entries and ITAB2 has n2 entries, the time needed for
    the nested loop with the straightforward algorithm is O(n1 * n2),
    whereas the parallel cursor approach takes only O(n1 + n2) time.
    The above parallel cursor algorithm assumes that ITAB2 contains only
    entries also contained in ITAB1.
    If this assumption does not hold, the parallel cursor algorithm
    gets slightly more complicated, but its performance characteristics
    remain the same.

    However for the code given here it is O (n1*log n2). Surely its a far more efficient algorithm
    compared to original nested loop with O (n1*n2) runtime. BUT, I am nNot sure if this is
    the intended parallel cursor algorithm.  which is supposedly O (n1 + n2)