Update query combined by group and set minimum dat

Question

I'm totally confused about how to get this query done.

How to achieve this query with no big performance impact (if table contains thousands rows) ?

Please advice!

MyTable

pktimestmp | sw_id | url_id |  open_date   | end_date  
-------------------------------------------------------
xxx0       |  101  |  com   |  2013-01-01  | 2013-01-30
xxx1       |  202  |  net   |  2013-01-01  | 2013-01-30
xxx2       |  202  |  net   |  2013-01-15  | 2013-02-28 *
xxx3       |  303  |  com   |  2013-01-01  | 2013-01-30 
xxx4       |  303  |  com   |  2013-02-01  | 2013-02-15 *
xxx5       |  303  |  com   |  2013-03-01  | 2013-03-15 *
xxx6       |  404  |  org   |  2013-01-01  | 2013-01-30 
xxx7       |  404  |  org   |  2013-02-01  | 2013-02-15 *
xxx8       |  404  |  gov   |  2013-02-01  | 2013-02-15
xxx9       |  404  |  gov   |  2013-02-01  | 2013-02-15 

...

update query on:

dates where software and url are the same (i.e. sw_id and url_id).

We set opendate and enddate to less (min) occurrence.

xxx2       |  202  |  net   |  2013-01-15  | 2013-02-28
will be
xxx2       |  202  |  net   |  2013-01-01  | 2013-01-30 

here we have updated 
open_date to min(open_date)
and 
end_date to min(end_date)
from same sw and url.

(it would be nice if we only update where there is a difference, for example no need to update xxx8 and xxx9 as they are "same")

MyTable ( * = updated rows )

pktimestmp | sw_id | url_id |  open_date   | end_date
-------------------------------------------------------
xxx0       |  101  |  com   |  2013-01-01  | 2013-01-30
xxx1       |  202  |  net   |  2013-01-01  | 2013-01-30
xxx2       |  202  |  net   |  2013-01-01  | 2013-01-30 *
xxx3       |  303  |  com   |  2013-01-01  | 2013-01-30 
xxx4       |  303  |  com   |  2013-01-01  | 2013-01-30 *
xxx5       |  303  |  com   |  2013-01-01  | 2013-01-30 *
xxx6       |  404  |  org   |  2013-01-01  | 2013-01-30 
xxx7       |  404  |  org   |  2013-01-01  | 2013-01-30 *
xxx8       |  404  |  gov   |  2013-02-01  | 2013-02-15
xxx9       |  404  |  gov   |  2013-02-01  | 2013-02-15 

Thanks in advance! (And no I have no rights to re-model the table) :)

Any help is appreciated, urls, hints, ebooks, sites...

Answer 1

The following should work (on pretty much any RDBMS):

UPDATE MyTable as a SET (open_date, close_date) 
                        = (SELECT MIN(open_date), MIN(close_date)
                           FROM MyTable as b
                           WHERE b.sw_id = a.sw_id
                             AND b.url_id = a.url_id)
WHERE EXISTS (SELECT *
              FROM MyTable as b
              WHERE b.sw_id = a.sw_id
                AND b.url_id = a.url_id
                AND (b.open_date < a.open_date OR b.close_date < a.close_date))

Which generates the expected, with only 4 rows updated (tested against my iSeries instance):

PKTIMESTAMP         SW_ID  URL_ID  OPEN_DATE   CLOSE_DATE
xxx0                101    com     2013-01-01  2013-01-30
xxx1                202    net     2013-01-01  2013-01-30
xxx2                202    net     2013-01-01  2013-01-30
xxx3                303    com     2013-01-01  2013-01-30
xxx4                303    com     2013-01-01  2013-01-30
xxx5                303    com     2013-01-01  2013-01-30
xxx6                404    org     2013-01-01  2013-01-30
xxx7                404    org     2013-01-01  2013-01-30
xxx8                404    gov     2013-02-01  2013-02-15
xxx9                404    gov     2013-02-01  2013-02-15