Re: ALTER in procedure

23 Jul 2013

      Seems like I found the root cause hopefully. Replaced DISTINCT with 
GROUP BY and it is working beautifully (in 15 secs).
It looks more like mindset shift. :)
Radovan

On 07/23/2013 01:12 PM, Stefan Manegold wrote:
...
Did you try to compare *the same* query between plain SQL and in procedure?
Mind in particular
SELECT COUNT(DISTINCT p.cislo_karty)
vs.
SELECT DISTINCT p.cislo_karty
(and I assume you use the same data and same timestampes in both/all cases)
If both of the above are fast in plain SQL but slow in a procedure,
you might want to use stethoscope [1] to find out where things hang.
Stefan
[1] http://www.monetdb.org/Documentation/Manuals/MonetDB/Profiler/Stethoscope
----- Original Message -----
...
Hi,
I looks like different kind of issue. I'm running into problem when
almost the same SQL running separately takes 5 seconds and in a
procedure it never finishes (30 minutes ...).
Following insert takes 5 seconds:
INSERT INTO crmkartastitek (
        cislo_karty,
        stitek_id,
        stitek_nazev,
        stitek_upresneni
       )
SELECT COUNT(DISTINCT p.cislo_karty),120,'sdfgdf','sdfgsdf'
       FROM crmkartaaktivita k
       JOIN crmkartapohyby p ON p.cislo_karty=k.cislo_karty
       WHERE 1=1
         AND k.aktivita > 0
         AND p.pohyb = 'VV'
         AND p.datum BETWEEN (timestamp'2013-07-01 00:00:00' - INTERVAL
'12' MONTH) AND (timestamp'2013-07-01 00:00:00' - INTERVAL '1' DAY)
Following procedure never returns:
DROP PROCEDURE crm_segm_kvv_rl
/
-- vyhodnotí pravidlo pro štítek KKV
-- @param p_interval dle číselníku crmregstitek
-- @param p_datum timestamp, první den v měsíci, např. timestamp
'2013-07-01 00:00:00'
CREATE PROCEDURE crm_segm_kvv_rl(p_interval INT, p_datum timestamp)
-- $Id: crm_segm_kvv_rl.sql 958 2013-07-22 08:40:30Z rbiciste $
BEGIN
    -- hlavička procedury
    DECLARE v_proc VARCHAR(250);
    SET v_proc = 'crm_segm_kvv_rl';
    -- nastav id pravidla
    DECLARE v_stitek_id int;
    DECLARE v_stitek_nazev,v_stitek_upresneni VARCHAR(200);
    -- nastav štítek
    SET v_stitek_id = 100;
    SELECT nazev,upresneni INTO v_stitek_nazev,v_stitek_upresneni FROM
crmregstitek WHERE stitek_id = v_stitek_id;
    -- proveď výpočet pouze pokud spouštím měsíční přepočet
    IF p_interval = crm_segm_vrat_spusteni(v_stitek_id) THEN
      INSERT INTO crmkartastitek (
        cislo_karty,
        stitek_id,
        stitek_nazev,
        stitek_upresneni
       )
       SELECT DISTINCT
p.cislo_karty,v_stitek_id,v_stitek_nazev,v_stitek_upresneni
       FROM crmkartaaktivita k
       JOIN crmkartapohyby p ON p.cislo_karty=k.cislo_karty
       WHERE 1=1
         AND k.aktivita > 0
         AND p.pohyb = 'VV'
         AND p.datum BETWEEN (p_datum - INTERVAL '12' MONTH) AND (p_datum
- INTERVAL '1' DAY);
    END IF;
END;
/
-- tests
call crm_segm_kvv_rl(1,timestamp '2013-07-01 00:00:00')
/
I tried:
1) comment out IF
2) replace variables v_xxx for constants like 'dfghdfgh' etc
Maybe I'm trying to do things Oracle way... but have not learned the
monet way yet.
Thank you,
Radovan
On 07/23/2013 10:23 AM, Stefan Manegold wrote:
...
Be aware that MonetDB will check/validate the constraints in either case,
i.e., either during loading (copy into ?) in case constraints are enable,
or once you re-enabled them after loading with constraints disabled.
Did you test the speed difference between loading with constraints enabled
vs. loading with constraints disabled plus enabling (checking) the
constraints afterwards?
(you can do this "by hand" in plain SQL, i.e., outside any procedure)
If so, what are your findings?
...
But maybe we
should use different approach like indexing. We have started to use
MonetDB recently.
We have a table hundreds millions of rows and we are tagging data in it.
Example SQL:
       INSERT INTO crmkartastitek (
         cislo_karty,
         stitek_id,
         stitek_nazev,
         stitek_upresneni
        )
        SELECT DISTINCT
        cislo_karty,v_stitek_id,v_stitek_nazev,v_stitek_upresneni
         FROM (SELECT cislo_karty
           FROM crmkartapohyby p
           JOIN crmkartaaktivita k ON p.cislo_karty=k.cislo_karty
           WHERE 1=1
            AND k.aktivita > 0
            AND p.pohyb IN ('VR','VV')
            AND p.datum BETWEEN (p_datum - INTERVAL '12' MONTH) AND
            (p_datum - INTERVAL '1' DAY)
           GROUP BY p.cislo_karty
           HAVING SUM(castka_r + castka_d) > SUM(castka_v)
         ) AS sumy_castek_pohybu;
What exactly it the problem (if any)?
Stefan
_______________________________________________
users-list mailing list
users-list@monetdb.org
http://mail.monetdb.org/mailman/listinfo/users-list
-- 
__________________________

Radovan Bičiště
ceos data s.r.o.
třída SNP 402/48
500 03 Hradec Králové
Czech Republic

mobil CZ: +420 601 563 014
skype: rbiciste