Parallel ByLine, ByChunk?

[dsimcha]

On 8/10/2011 7:58 PM, Graham Fawcett wrote:
> On Tue, 09 Aug 2011 20:22:40 +0000, dsimcha wrote:
>
>> Ok, I've actually managed to come up with a good way to do this without
>> writing any new low-level concurrency code.  The idea is to create two
>> delegates:  nextDel reads the next element into a user-provided buffer.
>> emptyDel tests whether there's anything left to read.
>
> Doesn't that essentially implement the InputRange protocol, but with
> delegates instead of an object?
>
> Not saying it's bad, just wondering what the benefits/tradeoffs of your
> delegate-based approach are.
>
> Graham

The fundamental difference between what I've defined and an input range 
is that the next() delegate lets you provide your own buffer.  Of 
course, you can implement input ranges on top of this by defining an 
object that manages the buffering, which is exactly what I did.

I could have used objects instead (e.g. a class/struct with a next() and 
empty() member function), but as a matter of personal preference I err 
on the side of avoiding objects because, when used excessively, they 
lead to boilerplate code.