Re: select for update question

>> I'm looking into how to update a row in a table while protecting it
>> from access by others.
>>
>> so far I've come up with this solution:
>>
>> create function dostuff() returns void as $$
>> declare
>>   tmp integer;
>> begin
>>   select id into tmp from tableX where id>305 limit 1 for update;
>>   update tableX set some_field = some_value where id=tmp;
>> end; $$ language plpgsql;
>>
>> will that guarantee that  the row I selected wil be updated within
>> this function and no one else can sneak in between and update or
>> delete the row?
>>
>>
>> What would I use if I would write
>>
>> lock table   tableX  IN .... MODE  at the start of my function?
>>
>> Any particular benefit with either method?
>
> If you lock the table, the whole table is locked. The first method (with
> select for update) locks only the one record you want to update.
>
> For real multi-user-access the first method are better.

When will it stop beeing a better method? When you select a large
enough percentage of the rows?