Re: Help with NSPointerArray and NSMapTable

On Sep 23, 2009, at 10:42, Jonathan Dann wrote:

> On 23 Sep 2009, at 01:27, Christiaan Hofman wrote:
>
>>
>> On Sep 23, 2009, at 1:25, Jonathan Dann wrote:
>>
>>>
>>> On 23 Sep 2009, at 01:10, Christiaan Hofman wrote:
>>>
>>> [snip]
>>>
>>>> Right, that's basically what I proposed as well. So I also need  
>>>> to use this for simple types like CGFloat that are not pointer- 
>>>> sized?
>>>
>>> Yep.
>>>
>>> [snip]
>>>
>>>> Nice samples, that surely helps. Thanks. I just find it confusing  
>>>> that for NSInteger it works differently. Though I guess I could  
>>>> do the same.
>>>
>>> No problem. When you're doing this, you're not casting an integer  
>>> to a pointer, so AFAICT the caveats in the 64-bit docs don't apply  
>>> here.
>>>
>>> Jonathan
>>
>> As I understand it now, you do cast an integer to a pointer when  
>> you use NSPointerFunctionsIntegerPersonality. So these caveats do  
>> apply.
>
> Yeah but you don't have to use the integer personality with integers  
> or floats. If you have a struct with a single int then you can case  
> between the struct and an int without loss of precision.
> I'm pretty sure that's what we're doing here. See:
>

But would you have to use the intermediate struct? Can't I just use  
e.g. a CGFloat (passed in by address) directly using  
NSPointerFunctionsStructPersonality? As I understand it this option  
does not really care about the exact data type, only a blob of memory  
of a certain size.

But I was (also)  talking about directly adding NSInteger with  
NSPointerFunctionsIntegerPersonality.

Christiaan

> http://www.algorithm.com.au/downloads/talks/objective-c-internals/objective-c-internals.pdf
>
>> Christiaan
>>
>