=?UTF-8?B?Ik5vcmRsw7Z3Ig==?= <per.nordlow gmail.com> writes:

Assumed I have the following code

     SysTime[] times;
     const n = 3;
     foreach (i; 0..n) times ~= Clock.currTime;

is there a simpler, perhaps functional, higher order pattern with 
which to achieve the same goal?

"Tobias Pankrath" <tobias pankrath.net> writes:

On Saturday, 22 February 2014 at 12:29:11 UTC, Nordlöw wrote:
 Assumed I have the following code

     SysTime[] times;
     const n = 3;
     foreach (i; 0..n) times ~= Clock.currTime;

 is there a simpler, perhaps functional, higher order pattern 
 with which to achieve the same goal?

What's Clock.currTime exactly, a function?
--
times = iota(3).map!(x => Clock.currTime).array;
--

=?UTF-8?B?Ik5vcmRsw7Z3Ig==?= <per.nordlow gmail.com> writes:

On Saturday, 22 February 2014 at 12:40:42 UTC, Tobias Pankrath 
wrote:
 On Saturday, 22 February 2014 at 12:29:11 UTC, Nordlöw wrote:
 Assumed I have the following code

    SysTime[] times;
    const n = 3;
    foreach (i; 0..n) times ~= Clock.currTime;

 is there a simpler, perhaps functional, higher order pattern 
 with which to achieve the same goal?

 What's Clock.currTime exactly, a function?
 --
 times = iota(3).map!(x => Clock.currTime).array;
 --

Great!

See also: 
https://stackoverflow.com/questions/21954381/range-construction-pattern/21954416?noredirect=1#21954416

"Tobias Pankrath" <tobias pankrath.net> writes:

 Great!

 See also: 
 https://stackoverflow.com/questions/21954381/range-construction-pattern/21954416?noredirect=1#21954416

I don't think that 'create' is a good name for a function, that 
basically lazily evaluates another function foo n times. Maybe 
std.range.repeat should have an overload that takes an alias to a 
callable, so you don't need this rather awkward use of iota.

=?UTF-8?B?Ik5vcmRsw7Z3Ig==?= <per.nordlow gmail.com> writes:

 times = iota(3).map!(x => Clock.currTime).array;
 --

Ok here's my try so far:

auto create(alias fun)(size_t n)
{
     import std.range: iota, map;
     return n.iota.map!(n => fun);
}

Now what remains is to retstrict create to only take a fun with 
*no* input arguments and a non-void return.

How do I do that?

Philippe Sigaud <philippe.sigaud gmail.com> writes:

 Now what remains is to retstrict create to only take a fun with *no*

input arguments and a non-void return.
 How do I do that?

With a template contraint.
Not tested:

import std.traits: isCallable, ParameterTypeTuple, ReturnType;

auto create(alias fun)(size_t n)
if (isCallable!fun && ParameterTypeTuple!(fun).length == 0
&& !is(ReturnType!fun == void))
{
    import std.range: iota, map;
    return n.iota.map!(n => fun);
}

=?UTF-8?B?Ik5vcmRsw7Z3Ig==?= <per.nordlow gmail.com> writes:

On Saturday, 22 February 2014 at 14:03:11 UTC, Philippe Sigaud 
wrote:
 Now what remains is to retstrict create to only take a fun 
 with *no*

 input arguments and a non-void return.
 How do I do that?

 With a template contraint.
 Not tested:

 import std.traits: isCallable, ParameterTypeTuple, ReturnType;

 auto create(alias fun)(size_t n)
 if (isCallable!fun && ParameterTypeTuple!(fun).length == 0
 && !is(ReturnType!fun == void))
 {
     import std.range: iota, map;
     return n.iota.map!(n => fun);
 }

Your solution with

    ParameterTypeTuple!(fun).length == 0

doesn't work here because currTime takes a defaulted argument.

I believe we need a new std.trait, say arityMin, for this.
Any ideas on how to implement that? Perhaps using 
__traits(compiles...?

See also my update at:
https://stackoverflow.com/questions/21954381/range-construction-pattern/21954416?noredirect=1#21954416

"Tobias Pankrath" <tobias pankrath.net> writes:

 I believe we need a new std.trait, say arityMin, for this.
 Any ideas on how to implement that? Perhaps using 
 __traits(compiles...?

 See also my update at:
 https://stackoverflow.com/questions/21954381/range-construction-pattern/21954416?noredirect=1#21954416

In the meantime you can use an is-expression:
--
int foo()  { return 12; }
void bar() { }
void baz(int x) { }
void main()
{
	pragma(msg, is(typeof(foo()) == void));
	pragma(msg, is(typeof(bar()) == void));
         pragma(msg, is(typeof(baz()) == void));
}
--

You can check if arbitrary code compiles with the 
"if-typeof-delegate-trick":
--
static if(is(typeof({ <code>}))))
{
// code compiles
}
--

Philippe Sigaud <philippe.sigaud gmail.com> writes:

 You can check if arbitrary code compiles with the

"if-typeof-delegate-trick":
 --
 static if(is(typeof({ <code>}))))
 {
 // code compiles
 }
 --

Or with __traits(compiles, <code>), which better documents the intent.

=?UTF-8?B?Ik5vcmRsw7Z3Ig==?= <per.nordlow gmail.com> writes:

My try so far:

import std.traits: isCallable, ReturnType, arity, 
ParameterTypeTuple;

enum arityMin0(alias fun) = __traits(compiles, fun()); // new 
syntax in 2.064

auto repeat(alias fun)(size_t n) if (isCallable!fun &&
                                      arityMin0!fun &&
                                      !is(ReturnType!fun == void))
{
     import std.range: iota, map;
     return n.iota.map!(n => fun);
}

Philippe Sigaud <philippe.sigaud gmail.com> writes:

per.nordlow:
 My try so far:

 import std.traits: isCallable, ReturnType, arity, ParameterTypeTuple;

 enum arityMin0(alias fun) = __traits(compiles, fun()); // new syntax in

2.064
 auto repeat(alias fun)(size_t n) if (isCallable!fun &&
                                      arityMin0!fun &&

                                      !is(ReturnType!fun == void))
 {
     import std.range: iota, map;
     return n.iota.map!(n => fun);
 }

For a nullary function, this solution is certainly OK: short and
understandable.
I found a way to parse an entire function param list, including default
args, but I don't remember right now how I did that. Maybe by parsing
'fun.stringof'
Maybe this exists in Phobos now?

btw, I'd call this template 'apply', because repeat already exists in
Phobos with a different use.

=?UTF-8?B?Ik5vcmRsw7Z3Ig==?= <per.nordlow gmail.com> writes:

 btw, I'd call this template 'apply', because repeat already 
 exists in
 Phobos with a different use.

Good idea! I'll rename it.

I have a pile of extensions to std.algorithm, std.traits, 
std.numeric, I'll pull some day when I get the time...

/Per

=?UTF-8?B?Ik5vcmRsw7Z3Ig==?= <per.nordlow gmail.com> writes:

 I have a pile of extensions to std.algorithm, std.traits, 
 std.numeric, I'll pull some day when I get the time...

I mean push...