Join Performance in Detail

Question

Hi all,
I have a special question for the performance of the join statement. Can anyone explain me how the join really works. If I have a SELECT Statement like the following:
SELECT MARA~MATNR
       MARC~WEKRS
  FROM MARA JOIN MARC
    ON (MARAMATNR = MARCMATNR)
WHERE MARC~MMSTA = 01
ORDER BY MARC~WERKS.
In my example MARC has 1.000.000 entries and MARA has 200.000 . Can someone explain which is done first? I think first he made the join means, he creates a temp table with 1.000.000 entries, when delete all that do not match the SELECT Clause and than make the ORDER BY. But I am not really sure, so maybe someone can help me.
Thanks,
  Chris

Answer 1

Make a report with your select and check the time elapsed, then repeat changing the order of your tables in the FROM section, it will answer your question.
IMO, it will depend of your DB server. But with the few things I learned about SAP and ABAP, I think it will be faster (more memory needed, but) using internal tables.
Fill your itab_marc first, filtering the 1000k registers and loop it. If you do the insertions in your itabs using ORDER BY clause SQL, the result will be fast.
Maybe it doesn't help, but if you do it, post the results
Regards,
Vic
SELECT *
  FROM MARC
  INTO TABLE it_marc
  WHERE MMSTA = 01
  ORDER BY MATNR.
SELECT *
  FROM MARA
  INTO TABLE it_mara
  ORDER BY MATNR.
loop it_marc.
  loop it_mara.
    if it_mara-matnr = it_marc-matnr.
    endif.
  endloop.
endloop.
My doubt: if you use nested loops... the nested one will be restarted for each iteration of the first one??
If it is, you can improve the speed making a delete in it_mara where matnr < it_marc-matnr.