Re: Getting the index for an element in mutually referencing containers

["Alejandro R. Mosteo" <alejandro@mosteo.com>]

On 12/03/17 09:22, Simon Wright wrote:
> Niklas Holsti <niklas.holsti@tidorum.invalid> writes:
>
>> On 17-03-11 23:46 , Simon Wright wrote:
>>
>>> Basically, I (Simon, now) am having trouble thinking of an
>>> application where reference counting would be an appropriate
>>> solution.
>>
>> How about this one: the SW generates messages (telemetry packets)
>> reporting various sorts of data and events.
>>
> [...]
>>
>> Each destination has a queue of incoming messages, and the queueing
>> and processing time for a given message varies accordingly. Rather
>> than copy the (possibly long) message into each destination's queue,
>> the queues hold references to a single, shared instance of the
>> message, dynamically allocated in a memory pool. These references are
>> counted. When, finally, all destinations have processed the message,
>> the message's reference count reaches zero, and the message can be
>> deallocated.
>>
>> This use of reference counting is a typical design in satellite
>> on-board SW.
>
> Thanks for the example!
>
> I still have a feeling that this is an in-computer software technique
> which resolves a software problem, rather than a necessary response to
> an application-domain problem.

This is how (most? I can only vouch for Java) garbage-collecting 
languages work. Thus is a common way of thinking for many programmers (I 
would say most think first of ref-counted ptrs before weak refs when 
faced with the problem being discussed, but that's just my impression). 
And the source of a kind of memory leak, which is the problem you get in 
exchange for no dangling pointers...

Not surprisingly, C++ has standardized the whole shebang, with both 
semantics being discussed here covered:

http://en.cppreference.com/w/cpp/memory

Alex.