buffered input

[Andrei Alexandrescu]

On 2/6/11 4:13 PM, Nick Sabalausky wrote:
> "Andrei Alexandrescu"<SeeWebsiteForEmail at erdani.org>  wrote in message
> news:iimnm6$1m4a$2 at digitalmars.com...
>> On 2/6/11 12:42 PM, spir wrote:
>>> Also: could a (truely) circular buffer help&  solve the above copy
>>> problem, concretely?
>>
>> Not if you want infinite lookahead, which I think is what any modern
>> buffering system should offer.
>>
>
> Agreed, but not everything always needs infinite lookahead. And sometimes
> space guarantees are important.

If you don't use infinite lookahead you won't consume infinite memory. 
You stand to consume more memory though but I believe we're not supposed 
to optimize for that.

To implement O(n) buffering for n elements of lookahead you can use the 
circular buffer already present in std.range as backend, coupled with 
routines to replenish it. The disadvantage is that the client doesn't 
see true T[] buffers, it sees lookalike random-access ranges that use 
modulo operations for indexing. All in all this is an abstraction worth 
providing but not as general as discardFromFront() and appendToFront().


Andrei