Mailing Lists: Apple Mailing Lists
Image of Mac OS face in stamp
 
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Memory Management (Part I)

original posting is stuck, coz it's to big, so here it's again in two parts..
enjoy or suffer :)


[I don't comment on everything, where I think you're just rewording what I wrote :)]

Am Mittwoch, 30.07.03 um 22:16 Uhr schrieb Marcel Weiher:

In other environments the assignment would prevent result from being freed during the lifetime of the variable. Since it cannot be done here the "solution" is too extend a little bit the lifetime of objects so that it works a little bit more like a garbage collected environment.

This is thinking in another language, and forcing Foundation / Objective-C into that mindframe. It doesn't help putting "garbage collection" as a concept or a goal to be aspired to, into this mail thread.

Where on earth did you get that from?!? Reference counting is a form of garbage collection. It isn't a perfect form of garbage collection because it can't handle cycles, but it still is a form of garbage collection. Retain/release, of course, is not fully automatic garbage collection, because it is at best semi-automatic.

The previous poster states that it is done to "work a little bit more like a garbage collected environment", then I am saying: Don't think about garbage collection, just think in terms of autorelease pools. It doesn't help to expand the domain, then get certain amounts of expectations going, that it oughta be doing this and that and then be disappointed, when it fails. Read the sentence in context with the previous discusssion, it will make sense.




Anyway, all this is only necessary when the callee does not hang on to the object. If the callee does hang on to the object, there simply is no *need* to extend the life of the object.

[examples]

Autorelease pools have some clear semantics. Every object added to an autorelease pool gains a property, a guaranteed minimum lifetime. Not more not less. Autorelease therefore guarantees (if result is autoreleased) that your case 1 will work

It guarantees no such thing.

See my other post just before this one. It does guarantee it.


Otherwise I am really curious how you are going to "break" this

- doSomething
{
result = [ofProcessObject processedString];
[ofProcessObject release];
[self doSomethingElse];
}

- doSomethingElse
{
NSLog (@"The result is %@", result);
}

when processedString returns with "retain] autorelease]".



and also guarantees that case 2 has a high chance of failure.

Ditto.

Well lets do case 1 first :)



What I'd like to add is this. Sharing a return value between caller and method ideally should be fail safe.

There are several levels to this statement.

At the most general level you are asking for referential transparency, something you are never ever going to get in an imperative language, and Objective-C is always going to remain an imperative language.

However, I think it is very good that you did bring it up in this generality, because I think that this is the (somewhat hidden) motivation behind some of the things that are being said. I find this motivation understandable, because referential transparency is a nice thing to have. However, the only way to actually get referential transparency is to have a language designed from the ground up to support it. It is one of these things that you absolutely cannot bolt on to a language after the fact. How this usually works is that there are *no* references whatsoever, only values. So there can never be aliasing of pointers/references because there are no pointers/references. Conceptually, everything is copied all the time. So if you insert an element into a collection, a new collection is created that has the element inserted at the right spot. The old collection is not modified. Nothing is ever modified.

I ain't asking for nothing here :), I am pretty much completely happy with Objective-C as it is, because I can achieve what you call "referential transparency" _if_ _I_ _want_ _to_. But in many cases I am very happy that the language doesn't force me to do that. That's why you and me aren't coding Java (except for money) :)


For those int, double return values it's trivial, because the value is copied through registers or stack. If there is memory involved then there needs to be a clear protocol for the transfer or sharing of that memory.

It isn't really the memory that is a problem. It is references to values (vs. values) and the aliasing that can happen when you are allowed to have more than one reference to a value.

This is just the same. The references always point to memory. Within that memory there can be references to other memory areas (*). If memory is deeply copied and owned by the caller then he has full ownership. There is nothing added by calling the memory object.


If the memory is owned by the caller, then there is no problem, the caller temporarily grants read/write access to its memory area, and can peruse it at will on return:

- (void) getName:(NSMutableString *) s
{
[s setString:name_];
}

Yes. However, I think you are forgetting that we are dealing with objects. Just as it is valid for one method to grant access to its 'memory area' to another method, it is just as valid for one *object* to grant access to its 'memory area' (instance variables) to another object.

I am afraid you don't follow my argument here. I was at that paragraph exactly just talking about the caller providing the memory, not the other way round (that's for later). Calling it object instead of memory doesn't make it any different, more complex maybe, yes.


this is ideal and fail safe. It is not done, because it is deemed to cumbersome. The caller must create temporary objects and release >> them.

It is not fail-safe, but it is cumbersome.

According to my discussion, it is fail safe. The memory area was provided for the method as a return value. If the method chooses to retain it, it's breaking the contract. It's equally breaking the contract if it does memset( s, 0, 16) on it.


If the memory is solely owned by the method (instance), this means a weakening in the fail-safeness

I don't agree with this. It is sharing, just like your other case, except that it is the *object* doing the sharing, not the method.

Well looking from the caller (and that is the way I am looking at it in this discussion) it is, because the caller must know that the return value is tied to the life and state of the called instance. (see below)



- (NSString *) name
{
return( name_);
}

This transfers a pointer to memory, not owned by the caller.

In the first case, you also transfered a pointer to memory, from the caller to the callee. This is just as unsafe because the callee could have released the memory. There is no difference!

Yeah could have, but that's "make pretend". The callee could also memset the stack...
See how neatly the first case avoids the problem here:

- doSomething
{
NSMutableString result;

result = [[NSMutableString alloc] init];
[ofProcessObject getProcessedString:result];
[ofProcessObject release];
[self doSomethingElse:result]; // no problemo
[result release]; // caller has full control
}


vs. the shared case

- doSomething
{
NSString *result;

result = [ofProcessObject processedString];
[ofProcessObject release];
[self doSomethingElse:result]; // oh oh, invalid
}




------------------------------------------------------
Ich selbst habe mich persvnlich stets an den
sittlichen Gepflogenheiten des Philisters beteiligt;
nicht, weil ich ihnen Wert beigemessen hdtte, sondern
weil ich - vorwdrts, immer mal wieder! - keinen Anla_
traf, mich von ihnen zu trennen. -- H. Mann
_______________________________________________
cocoa-dev mailing list | email@hidden
Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/cocoa-dev
Do not post admin requests to the list. They will be ignored.


Visit the Apple Store online or at retail locations.
1-800-MY-APPLE

Contact Apple | Terms of Use | Privacy Policy

Copyright © 2007 Apple Inc. All rights reserved.