On Mon, 15 Mar 2010, Chris Muller wrote:

>> On the other hand, current stupid implementation prevents performing a
>> useless copy, except near end of stream...
>
> Yes, that's a big problem!  Not only do I no longer have access to the
> primitive, it costs me an allocation every time I want just to read
> from a file into MY ByteArray.  Aren't there any other solutions?  Can
> we talk about this?
>
>

Thanks for the bug report Chris and for fix Nicolas. I think we now have 
the best we can. I doubt that losing direct access to the primitive 
will be a bottleneck. On my pc, I could achieve 360MB/second read speed 
with the preivous implementation of #next:into:startingAt: if the data was 
already cached in memory. If you really have to access the primitive, you 
can still do that by turning off the read buffer (#disableReadBuffering) 
and use the primitive accessor #primRead:into:startingAt:count: directly.


Levente