Performance tuning the loop statement

[Former Member]

Hi,

I have the following requirement.

I have an internal table IT_TEST1 which has values

Column1

A

B

C

D

..

consider a second internal table IT_TEST2. It'll have values like as follows:..I need to know which are the entries in column1 that have more than one entry in column 2 without looping through this internal table..Kindly suggest..

Column1 column2

A VAL1

B VAL1

B VAL2

C VAL1

C VAL2

D VAL1

D VAL2

D VAL3

This has caused very big performance issue for it is

[Former Member]

Hi,

I would suggest do the read with binary serach over second table and then in addition check if thenext entry has the same key.

e.g.

READ ITAB2 WITH KEY k = A BINARY SEARCH

lv _idx = sy-tabix + 1.

READ ITAB2 INDEX lv _idx.

...check if the second record has the same key -- if so you have at least duplicates

BR, Artem

remember points

[Former Member]

Hi!

You can try this.

1. Make a temporary internal table and copy IT_TEST2 into this IT_TEMP.

2. Now LOOP AT IT_TEST1

3. READ TABLE IT_TEMP WITH KEY Column1 = IT_TEST1-column1 binary search.

4. DELETE IT_TEMP. This will delete the current entry from IT_TEMP

5. Now again read IT_TEMP. READ TABLE IT_TEMP WITH KEY Column1 = IT_TEST1-column1 binary search.

6. If SY-SUBRC = 0 then you have more than one entry in IT_TEST2.

7. ENDLOOP.

Reward points if useful.

[Former Member]

I would suggest ..

First sort the table by column1... so all records are in sorted form..

sort ITAB by column1..
read table ITAB with key column1 = 'A'.
l_index = sy-tabix + 1.

read table ITAB index l_index.

if sy-subrc eq 0.
write:/ 'More than 1 record'.
else.
write:/ 'Single record only'.
endif.

Reward if useful

Regards

Prax

[Former Member]

Hi,

I have faced a similar problem and the following strategy has saved a lot of processing time

the following code may serve your pupose

if you are using IT_TEST2 for further processing please use IT_TEST3

otherwise use IT_TEST2 instead of IT_TEST3 in the code below......'

*************************CODE SAMPLE************************************

data : IT_TEST3 like IT_TEST2. "declare internal table same as IT_TEST2

IT_TEST3[] = IT_TEST2[].

"Pass contents if IT_TEST2 is used for further processing

Sort IT_TEST1 by column1 . "Sort

Sort IT_TEST3 by column1 . "Sort

loop at IT_TEST1.

read table IT_TEST3

with key column1 = IT_TEST1-column1.

if sy-subrc = 0.

describle table IT_TEST3 lines n.

if lines > 1.

"Multiple entries

else.

"Single entry

endif.

endif.

"Perform all operations on IT_TEST3

DELETE IT_TEST3 WHERE column1 = IT_TEST1-column1.

"Delete the entry in IT_TEST3 already processed with the key of IT_TEST1

"Next time you do the read entries for the previous and all predecessor loops

will not be there'

"As you go down the loop performance increases

endloop.

Regards

Byju

[Former Member]

I think there is not one decent solution to your problem as all use nested loop which will kill your performance.

And I am not sure whether I really understand your problem. To count the number of entries make it necessary to loop over the whole table, therefore you must

loop of table2. But you should not loop inside another loop on table1.

I propose precprocessing:

define a hashed table itab3 with key column2 and count

loop at itab2 into wa2.

read itab3 with table key column2 = wa2-column2.

if ( sysubrc eq 0 )

wa3-count = wa3-count + 1.

modify itab3

else.

insert itab3

endif.

endloop.

=> itab3 will tell you how often the values in itab2 appear. It need only one loop.

Then it is very simple to solve your problem

loop itab1..

read itab3

if ( count > 1 ).

endif.

endloop

Coding is not exactly correct. Please correct, if I did not understand your problem.

Siegfried

Message was edited by:

Siegfried Boes

[Former Member]

Use the parallel cursor method of looping through the two tables: Read <a href="/people/rob.burbank/blog/2006/02/07/performance-of-nested-loops of Nested Loops</a>.

MattG.