[V4CC] about autorelease of Vcursor

[Ruslan Zasukhin]

On 8/17/10 6:08 PM, "Erne" <ernestogiannotta at tiscalinet.it> wrote:

Hi Erne,

>>> encountered some issues with this build:
>>> 
>>> first why did you remove the VCursor contructor (initWithDB)?
>> 
>> Because such constructor was only in V4RB some time ago,
>> And it was "copied" into  V4CC.
>> 
>> All other Valentina ADKs do not have it.
>> So we just  have to unify API.
>> 
>> 
>> --------------- 
>>> This was a lot useful when trying to query several cursor in a row with a
>>> readwrite lock
>> 
>> This is not clear to me. Example ?
>> 
>>> I mean the standard DB SqlQuery method returns an autoreleased object that
>>> will keep the lock alive for the entire main loop cycle.
>> 
>>> If you don't want we alloc init our cursors we at least need a destructor or
>>> unlock method for the cursor.
>> 
>> Again - explain please.
>> 
>> Just have point that  since V4CC have start to use autorelease.
>> Developers should start to use NSAutoReleasePool to control life of cursor.
>> 
> 
> Well, for the [mDB sqlSelect] method it's correct to return an autoreleased
> VCursor object 
> and that's OK for most tasks.
> 
> But imagine this simple scenario:
> I got a method that queries the db like this to populate list a
> 
> VCursor *c1 = [mDB sqlSelect:sqlString
> cursorLocation:EVCursorLocation_kServerSide
>   lockType:EVLockType_kNoLocks
>  direction:EVCursorDirection_kRandom];
> 
> now in the same method I call another method that eventually queries the same
> db to populate list b
> 
> VCursor *c2 = [mDB sqlSelect:anotherSqlString
> cursorLocation:EVCursorLocation_kServerSide
>   lockType:EVLockType_kNoLocks
>  direction:EVCursorDirection_kRandom];
> 
> If some record is already locked by the first query I get a Lock error by
> Valentina engine
> and there's no way for me to free that lock until I exit the first method and
> the autorelease pool will destroy c1 cursor
> because if I call [c1 release] before calling the second method
> then the autorelease pool will crash attempting to release the same object
> twice.
> 
> This will be most crucial in a multiuser environment where locks are most
> important
> 
> So it would either be useful a [c freeLocks] method
> or a contructor like we had before that will return a not autoreleased object
> so that we own our cursor and will be responsible for releasing it.

No no Erne,

Once again. Official way from Apple now is
    Use NSAutoReleasePool  !!

Foo()
{
    {
        NSAutoreleasePool *innerPool = [[NSAutoreleasePool alloc] init];

        VCursor *c1 = [mDB sqlSelect:sqlString
                     cursorLocation:EVCursorLocation_kServerSide
                           lockType:EVLockType_kNoLocks
                          direction:EVCursorDirection_kRandom];

        [ innerPool release];   <<<<< this will really destroy c1 !
    }


}


In the NET they also should use special construction to force true object
death: 

    using( Vcursor curs1 = db.SqlSelect(...) )
    {
    }


-- 
Best regards,

Ruslan Zasukhin
VP Engineering and New Technology
Paradigma Software, Inc

Valentina - Joining Worlds of Information
http://www.paradigmasoft.com

[I feel the need: the need for speed]