On 4/8/2015 10:33 AM, Adam D. Ruppe wrote:
On Wednesday, 8 April 2015 at 17:17:21 UTC, Walter Bright wrote:
writeln is an OutputRange.
No, it isn't. It isn't even remotely close to one - an OutputRange is a type,
I meant an instance of a type, as in:
i is an int
meaning i has type int.
writeln is a function. isOutputRange doesn't even return false for it, it
outright fails to compile!
I understand that a true OutputRange would require one to write:
arg.copy(writeln);
and that a common idiom is to write it so the copy() can be implicit. I haven't
made up my mind as to the right style here.
File isn't an output range either btw. An output range is defined by the
presence of a put method, which File doesn't have.
b.d(4): Error: static assert (isOutputRange!(File, char)) is false
Even putting aside the formal definition of output range, output is NOT part of
the pipeline! Pipelines can be composed and they keep pumping data through them.
An output range is the caboose of a pipeline (to mix metaphors).
writeln does not transform its input nor does it forward it to the next item in
the pipe. It just eats it.
It places its input into a container.
That's totally different than everything else in the sequence. And that's not
even bringing in writeln's sister function, writefln, whose first argument isn't
what is to be printed at all!
It's true that a lot of Phobos code was designed pre-ranges, and the parameters
aren't set up ideally for it. I've been doing some work to fix that.
