Thanks for your reply.
I would tend to agree with you but why doesn't it update ANY row when count() equals zero then, and ALL rows only when
thecount is not zero ? I would agree with you if ALL rows were updated in any case or did I miss something ?
Arnaud
>The first query SHOULD update all rows, because there is NO restriction
>on the utilisateur table - a.login <> '*' is NOT the same as
>utilisateur.login <> '*' - the former qualifies the FROM clause, and the
>latter applies to the UPDATE itself.
>"Dima Tkach" <dmitry@openratings.com> a écrit dans le message de news: 3CC6A7AF.3070801@openratings.com...
>> Here's the "wrong" query :
>> update utilisateur set login='*', password='*' from utilisateur a, ancien b where a.util_id=b.util_id and
b.anc_decesis not null and >(a.login<>'*' or a.password<>'*');
>>
>> Here's the "correct" one :
>> update utilisateur set login='*', password='*' from ancien b where utilisateur.util_id=b.util_id and b.anc_deces is
notnull and >(utilisateur.login<>'*' or utilisateur.password<>'*');
>>
>> Both queries work the same when there's nothing to update, (i.e. select count(*) from utilisateur a, ancien b where
>a.util_id=b.util_idand b.anc_deces is not null and (a.login<>'*' or a.password<>'*') returns 0).
>> But the first one ("wrong") updates ALL rows of table 'utilisateur' when the above count() is not 0, where the
secondone >("correct") updates only the rows needed.
>>
>> What's wrong with the first query : me or the parser ? I guess the problem comes from the fact that table
'utilisateur'appears twice and is aliased which may confuse the parser in the "wrong" query...
>>