Does REDUCE perform better using sorted tables?

[matteo_pascon]

So, I'm using reduce to get the number of record that satisfy a condition inside a loop.

Will it perform better if the table is declared as sorted using a logic similar to that of parallel cursor or does it actually just check linearly?

The code is as follow.

DATA(sum) = REDUCE i( INIT x = 0 FOR wa IN lt_iloa WHERE ( tplnr = ls_ifl_list-tplnr ) NEXT x = x + 1 ).

[matteo_pascon]

Hi Matteo,

Yes, it will take the advantage of access by key on tplnr and will not run entire internal table. So the difference of execution can be huge and with a constant result time benefit from indexed access.

Regards, Fernando Da Rós

[fedaros]

Hi Matteo,

Yes, it will take the advantage of access by key on tplnr and will not run entire internal table. So the difference of execution can be huge and with a constant result time benefit from indexed access.

Regards, Fernando Da Rós

[matteo_pascon]

Wonderful! Thanks for the insights.

[Sandra_Rossi]

matteo_pascon as documented in ABAP documentation FOR, cond > "As with LOOP AT itab, the following is possible for each expression FOR ... IN itab" > LOOP AT itab, cond :

• "If the primary table key is specified using its name primary_key, the processing behaves in the same way as if no key were explicitly specified."
• etc.

[matteo_pascon]

Got it. Thanks for sharing the documentation.