I liked the two preceding tips on performance improvement. I have some extra remarks. First of all, always be 100%...
sure that the table you use for FOR ALL ENTRIES is filled. Otherwise, a full table scan will be executed. My second remark is the usage of the sorted table TA_MARA and the READ TABLE statement. If you want to have optimal performance you either use READ TABLE ..BINARY SEARCH of READ TABLE WITH TABLE KEY. Using WITH KEY does not take advantage of the sorted table. If you are using a sorted table you do not have to use the logic with LOOP AT TA_MARA FROM TABIX. You can easily use the LOOP AT TA_MARA WHERE option (as long as you use the key fields of the sorted table). Finally, when using a table without table header, you have to use an implicit work area.
Report zzatip. Data: begin of data_package occurs 0, Matnr like mara-matnr, end of data_package. types: begin of ty_mara, matnr like mara-matnr, matkl like mara-matkl, end of ty_mara. data: wa_mara type ty_mara. data: ta_mara type sorted table of ty_mara with non-unique key matnr. * First, data must be available in the table DATA_PACKAGE * Then the corresponding materialdata from MARA must be read if not data_package is initial. SELECT MATNR MATKL INTO TABLE TA_MARA FROM MARA FOR ALL ENTRIES IN DATA_PACKAGE WHERE MATNR = DATA_PACKAGE-MATNR. endif. * With the following loop you can combine the records. *Let's complicate things and assume *dupplicates on ta_mara *(Could come from mard or marc) Data idx like syst-tabix. * No SORT TA_MARA needed, It is a sorted table! LOOP AT DATA_PACKAGE. read table ta_mara into wa_mara with TABLE key matnr = DATA_PACKAGE-MATNR. check syst-subrc = 0. idx = syst-tabix. Loop at ta_mara into wa_mara from syst-tabix. if wa_mara-matnr ne data_package-matnr. exit. endif. ENDLOOP. ENDLOOP. * Other option that works ok LOOP AT DATA_PACKAGE. Loop at ta_mara into wa_mara where matnr = data_package-matnr. ENDLOOP. ENDLOOP.