On Tuesday, 13 March 2012 at 14:31:59 UTC, Andrei Alexandrescu
wrote:
On 3/13/12 1:31 AM, Xinok wrote:
- I wrote it to sort random-access ranges *without* slicing,
but I think
the exclusion of slicing makes it slower. I'm writing a
separate
implementation which uses slicing and I'll keep it if it's
much faster.
Having random access implies having slicing.
Currently it can not be assumed that isRandomAccessRange has
slicing:
http://dlang.org/phobos/std_range.html#isRandomAccessRange
Maybe it should be a requirement?
It seems to me that Bidirectional ranges can't be infinite, and
by extension Random Access ranges too. But slicing could be
supported on an infinite range. So hasSlicing is still useful,
but I think could be a good requirement on RA ranges.
