Performance problem in ABAP programming

[jonathanzhang2]

Hi!

Please review the following program,

LOOP AT TG_PRICE INTO WA_PRICE.

CLEAR WA_VBELN.

READ TABLE TG_VBELN INTO WA_VBELN WITH KEY KNUMV = WA_PRICE-KNUMV

POSNR = WA_PRICE-KPOSN.

IF SY-SUBRC = 0.

WA_PRICE-VKORG = WA_VBELN-VKORG.

WA_PRICE-VBELN = WA_VBELN-VBELN.

WA_PRICE-POSNR = WA_VBELN-POSNR.

WA_PRICE-MATNR = WA_VBELN-MATNR.

WA_PRICE-MAKTX = WA_VBELN-MAKTX.

WA_PRICE-ERNAM = WA_VBELN-ERNAM.

WA_PRICE-ERDAT = WA_VBELN-ERDAT.

MODIFY TG_PRICE FROM WA_PRICE.

DELETE TABLE TG_VBELN FROM WA_VBELN.

ELSE.

DELETE TABLE TG_PRICE FROM WA_PRICE.

ENDIF.

ENDLOOP.

comments: internal table TG_PRICE has about 20,000 records and TG_VBELN has about 70,000.

this segment run spent too much time.

Comparing the following program, the speed is more faster.

LOOP AT TG_PRICE INTO WA_PRICE.

CLEAR WA_KONP.

SELECT SINGLE KNUMH

KBETR

KONWA

KPEIN

KMEIN

INTO WA_KONP

FROM KONP

WHERE KNUMH = WA_PRICE-KNUMH.

IF SY-SUBRC = 0.

WA_PRICE-KBETR_FROM = WA_KONP-KBETR_FROM.

WA_PRICE-KONWA_FROM = WA_KONP-KONWA_FROM.

WA_PRICE-KPEIN_FROM = WA_KONP-KPEIN_FROM.

WA_PRICE-KMEIN_FROM = WA_KONP-KMEIN_FROM.

VL_AMOUNT = WA_PRICE-KBETR_FROM.

CALL FUNCTION 'CURRENCY_AMOUNT_SAP_TO_DISPLAY'

EXPORTING

CURRENCY = WA_PRICE-KONWA_FROM

AMOUNT_INTERNAL = VL_AMOUNT

IMPORTING

AMOUNT_DISPLAY = VL_RATE

EXCEPTIONS

INTERNAL_ERROR = 1

OTHERS = 2 .

IF SY-SUBRC = 0.

WA_PRICE-PRICE_FROM = VL_RATE.

ENDIF.

MODIFY TG_PRICE FROM WA_PRICE.

ELSE.

DELETE TABLE TG_PRICE FROM WA_PRICE.

ENDIF.

ENDLOOP.

commmets: internal table TG_PRICE has also 20,000 records. Within the Loop, there is system table selection clause. it should be slower than the previous program. But why the previous is more slower, is there something wrong with the READ clasue or any others?

Many thanks,

Aries

[suresh_datti]

Hi Aries,

Use READ with BINARY SEARCH option to speed up the first Program and be sure to sort the itab with the key you are using in the READ. Also, the sy-subrc check should be before PSONR = WA_PRICE-KPOSN.

Regards,

Suresh Datti

[jonathanzhang2]

Hi Suresh,

PSONR = WA_PRICE-KPOSN is also a key for READ clause.

by the way, the internal tables have been sorted.

After using BINARY SEARCH, the speed is faster acually.

But still much slower than the second program LOOP with select clause. In my opinion, the operation with internal table should much faster than operation with select system table. Do you have any other ideas?

Many thanks,

Aries

[suresh_datti]

Hi Aries,

The second Program runs faster even if you are hitting the database for each loop pas due to the SELECT SINGLE. This statement has the complete key so retrieves faster. Can you pass the complete key with READ statement in the first program? may be that will show some improvement. Also,did you try out Andreas tip?

Regards,

Suresh Datti

[jonathanzhang2]

Hi Suresh,

The complete READ statment is as follows:

READ TABLE TG_VBELN INTO WA_VBELN

WITH KEY KNUMV = WA_PRICE-KNUMV

POSNR = WA_PRICE-KPOSN.

Hi Andreas,

The two internal tables have already been sorted as follows:

SORT TG_PRICE BY KNUMV KPOSN.

SORT TG_VBELN BY KNUMV POSNR.

And these two internal table are defined as follows:

• Price

TYPES:BEGIN OF YS_PRICE,

KNUMV TYPE KNUMV, "Number of the document condition

KPOSN TYPE KPOSN, "Condition item number

KNUMH TYPE KNUMH, "Condition record number

VKORG TYPE VKORG, "Sales orgnazation

VBELN TYPE VBELN_VA, "Order Number

POSNR TYPE POSNR_VA, "Item number

MATNR TYPE MATNR, "Material Number

MAKTX TYPE MAKTX, "Material Description

KBETR_FROM TYPE KBETR_KOND,"Changed From Price

KONWA_FROM TYPE KONWA, "Price Unit

KPEIN_FROM TYPE KPEIN, "Per

KMEIN_FROM TYPE KMEIN, "Per

KBETR_TO TYPE KBETR_KOND,"Changed to Price

KONWA_TO TYPE KONWA, "Price Unit

KPEIN_TO TYPE KPEIN, "Per

KMEIN_TO TYPE KVMEI, "Per

ERNAM TYPE ERNAM, "Created by

ERDAT TYPE ERDAT, "Date Created

PRICE_FROM TYPE P DECIMALS 2,

PRICE_TO TYPE P DECIMALS 2,

END OF YS_PRICE.

• Sales order

TYPES: BEGIN OF YS_VBELN,

KNUMV LIKE VBAK-KNUMV,"Number of the document condition

VKORG TYPE VKORG, "Sales orgnazation

VBELN TYPE VBELN_VA, "Order Number

POSNR TYPE POSNR_VA, "Item number

MATNR TYPE MATNR, "Material Number

MAKTX TYPE MAKTX, "Material Description

ERNAM TYPE ERNAM, "Created by

ERDAT TYPE ERDAT, "Date Created

END OF YS_VBELN.

DATA: TG_PRICE TYPE YS_PRICE OCCURS 0,

TG_VBELN TYPE YS_VBELN OCCURS 0.

Do you have any ideas?

Many thanks,

Aries

[andreas_mann3]

Hi,

make table TG_VBELN a <b>sorted</b> (and when unique - a <b>hashed</b> table )

-> there will be much better performance !

Andreas

[Former Member]

Hi for the second program make the changes as mentioned below.

1.dont use select inside the loop.

2. keep the select outside the loop.

then ur second program looks like this.

SELECT SINGLE KNUMH

KBETR

KONWA

KPEIN

KMEIN

INTO table WA_KONP

FROM KONP

for all entries in WA_PRICE

WHERE KNUMH = WA_PRICE-KNUMH.

sort wa_konp by knumh.

LOOP AT TG_PRICE INTO WA_PRICE.

CLEAR WA_KONP.

read table wa_konp with key tg_price-knumh = wa_konp-knumh binary search.

IF SY-SUBRC = 0.

WA_PRICE-KBETR_FROM = WA_KONP-KBETR_FROM.

WA_PRICE-KONWA_FROM = WA_KONP-KONWA_FROM.

WA_PRICE-KPEIN_FROM = WA_KONP-KPEIN_FROM.

WA_PRICE-KMEIN_FROM = WA_KONP-KMEIN_FROM.

VL_AMOUNT = WA_PRICE-KBETR_FROM.

CALL FUNCTION 'CURRENCY_AMOUNT_SAP_TO_DISPLAY'

EXPORTING

CURRENCY = WA_PRICE-KONWA_FROM

AMOUNT_INTERNAL = VL_AMOUNT

IMPORTING

AMOUNT_DISPLAY = VL_RATE

EXCEPTIONS

INTERNAL_ERROR = 1

OTHERS = 2 .

IF SY-SUBRC = 0.

WA_PRICE-PRICE_FROM = VL_RATE.

ENDIF.

MODIFY TG_PRICE FROM WA_PRICE.

ELSE.

DELETE TABLE TG_PRICE FROM WA_PRICE.

ENDIF.

ENDLOOP.

Satish

[Former Member]

Try to use the following code

FIELD-SYMBOLS <F_PRICE> LIKE WA_PRICE.
DATA : W_TABIX LIKE SY-TABIX,
       W_TABIX1 LIKE SY-TABIX.
LOOP AT TG_PRICE ASSIGNING <F_PRICE>.
  W_TABIX1 = SY-TABIX.
CLEAR WA_VBELN.
CLEAR W_TABIX.
READ TABLE TG_VBELN INTO WA_VBELN WITH KEY KNUMV = <F_PRICE>-KNUMV BINARY SEARCH.
IF SY-SUBRC = 0.
 W_TABIX = SY-TABIX.
ENDIF.

POSNR = WA_PRICE-KPOSN.

IF SY-SUBRC = 0.
<F_PRICE>-VKORG = WA_VBELN-VKORG.
<F_PRICE>-VBELN = WA_VBELN-VBELN.
<F_PRICE>-POSNR = WA_VBELN-POSNR.
<F_PRICE>-MATNR = WA_VBELN-MATNR.
<F_PRICE>-MAKTX = WA_VBELN-MAKTX.
<F_PRICE>-ERNAM = WA_VBELN-ERNAM.
<F_PRICE-ERDAT = WA_VBELN-ERDAT.
*MODIFY TG_PRICE FROM WA_PRICE. "NO NEED FOR THIS WITH
*FIELD SYMBOL 
*DELETE TABLE TG_VBELN FROM WA_VBELN.
 DELETE TABLE TG_VBELN INDEX W_TABIX.
ELSE.
DELETE TABLE TG_PRICE INDEX W_TABIX1.
ENDIF.
ENDLOOP.

I think it will make the code much faster because in your code three things are taking time.

(1) read without binary search

(2) Delete ( it traverse the whole table sequentially to find record to delete which has the same key values as in the header table

(3) It transports the data in header of the TG_PRICE table and then transport it back to the row to update it with modify statement. With 'assigning field symbol' you are directly working on row level

Please let me know how it works.

[jonathanzhang2]

Hi Sharad,

After using filed symbols, it is faster than before actually. Thanks.

Regards,

Aries