Re: [Pyobjc-dev] problem saving as rtfd

[Marcel W. <ma...@me...>]

On Monday, May 12, 2003, at 07:51  Uhr, Ronald Oussoren wrote:

>
> On Monday, May 12, 2003, at 08:43 Europe/Amsterdam, Marcel Weiher 
> wrote:
>
>>
>>>         plaintext = ""
>>>         cyphertext = NSData.alloc()
>>
>> Minor nit:  you shouldn't really have a partly initialized instance 
>> flying around.
>
> This is *not* a minor issue. Some Cocoa classes will cause crashes if 
> you do anything with them before calling init, and that includes 
> releasing them. PyObjC tries to protect you from this, but this is 
> still bad form.

Yes, actually doing anything with the object other than sending some 
sort of init message  *would* be bad.  However, the original poster 
didn't do that, and for the object in question, it actually makes no 
difference whatsoever, since only the sequencing matters (unless the 
code is seriously messed up).  It cannot tell what if anything happened 
between the alloc and the init.   It can be the very next thing, it can 
be a million years later.

	id object = [[NSObject alloc] init];
and
	id obj = [NSObject alloc];
	sleep( 365 * 24 * 3600 );
	obj=[obj init];

are completely equivalent as far as the object is concerned.  Of 
course, the latter is definitely odd looking, and there is no good 
reason for not initializing it immediately, whereas there are very good 
reasons *for* doing the initialization immediately, which is why I did 
bring it up.

Of course, with a bridge involved, you may have to be more careful 
because the bridge might have reasons for touching the object that 
aren't obvious from the code.

Marcel

-- 
Marcel Weiher				Metaobject Software Technologies
ma...@me...		www.metaobject.com
Metaprogramming for the Graphic Arts.   HOM, IDEAs, MetaAd etc.