data containers

[Steven Schveighoffer]

"Koroskin Denis" wrote
> On Fri, 04 Jul 2008 03:22:03 +0400, Jarrett Billingsley
>> "Moritz" wrote
>>> torhu schrieb:
>>>> Does it work if you make this change?
>>>>
>>>> -     if(layer <= layer_count)
>>>> +     if(layer < layer_map.length)
>>>
>>> Well, my program DOES work, but I doesnt do what I want it to, because 
>>> Im
>>> now unable to insert ScreenElements into the array.
>>>
>>> Do I need to initialize the array slots before I assign an object to 
>>> them?
>>> The way I see it, the array has length 0 at start, and I cant put 
>>> anything
>>> in slot 0 or Ill get an array out of bounds error :-(
>>
>> No, it's because you're using a temp variable when you shouldn't be.
>>
>> ScreenElement[] single_layer = layer_map[layer];
>> single_layer ~= new_element;
>>
>> That will get a reference the layer map at index `layer`, but when you
>> concatenate, only the local single_layer will refer to the newer, longer
>> array.  Just do it in place:
>>
>> layer_map[layer] ~= new_element;
>>
>> (Also if you don't feel like typing all those long types for variables, 
>> it's
>> completely unnecessary.
>>
>> auto single_layer = layer_map[layer];
>>
>> Dah.)
>>
>>> I guess my next problem after storing them will be to keep the garbage
>>> collector from deleting them after I leave the function where the
>>> ScreenElements  are created and stored into the array, or does the array
>>> reference protect the ScreenElements from GC?.
>>
>> Arrays are on the heap.  As long as something references said array, the 
>> GC
>> won't collect objects the array holds.  That's kind of the point.  The GC
>> isn't out to make your life hard; quite the opposite in fact :)
>>
>> The GC is actually fairly conservative, and if it's not sure whether
>> something is referenced or not (i.e. say you have some random-looking 
>> data,
>> like sound data, and the GC says "hm, this looks like a pointer into an
>> object"), it won't collect it.
>>
>>
>
> Or, better, use:
> auto single_layer = layer in layer_map;
>
> to avoid lookup on insertion:
> layer_map[layer] ~= new_element; -> single_layer ~= new_element;
>
> right?

According to an earlier post, layer_map is a 2D array, not an associative 
array.  I think 'in' only works on AA's.

What you could do though is:

auto single_layer = &layer_map[layer];

But I'm not sure ~ works on an array pointer, but D does do some magical 
stuff with operators when dealing with pointers, so it could work :)

-Steve