View mode: basic / threaded / horizontal-split · Log in · Help
August 21, 2010
map on fixed-size arrays
Hello,

The 'map' from std.algorithm doesn't seem to work with fixed-size arrays:

----------------------------------------------------------------------
import std.stdio ;
import std.math ;
import std.algorithm ;

T sq ( T ) ( T x ) { return x*x ; }

void main ()
{
  double [2] a = [ 1.0 , 2.0 ] ;

  writeln ( map ! ( sq ) ( a ) ) ;
}
----------------------------------------------------------------------

$ rdmd test_map_sq_fixed_size_b.d
/usr/include/d/dmd/phobos/std/algorithm.d(108): Error: template instance 
Map!(sq,double[2u]) does not match template declaration Map(alias 
fun,Range) if (isInputRange!(Range))

Is this an intended limitation?

Ed
August 21, 2010
Re: map on fixed-size arrays
Eduardo Cavazos:

> The 'map' from std.algorithm doesn't seem to work with fixed-size arrays:

See bug 4114
In my opinion map, sort, etc have to work with fixed-sized arrays too (otherwise I'll have to write more wrappers).

Bye,
bearophile
August 21, 2010
Re: map on fixed-size arrays
On 21.08.2010 14:37, Eduardo Cavazos wrote:
> Hello,
>
> The 'map' from std.algorithm doesn't seem to work with fixed-size arrays:
>
> ----------------------------------------------------------------------
> import std.stdio ;
> import std.math ;
> import std.algorithm ;
>
> T sq ( T ) ( T x ) { return x*x ; }
>
> void main ()
> {
>   double [2] a = [ 1.0 , 2.0 ] ;
>
>   writeln ( map ! ( sq ) ( a ) ) ;
> }
> ----------------------------------------------------------------------
>
> $ rdmd test_map_sq_fixed_size_b.d
> /usr/include/d/dmd/phobos/std/algorithm.d(108): Error: template 
> instance Map!(sq,double[2u]) does not match template declaration 
> Map(alias fun,Range) if (isInputRange!(Range))
>
> Is this an intended limitation?
>
> Ed
You always can workaround this by taking full slice:

import std.stdio ;
import std.math ;
import std.algorithm ;

T sq ( T ) ( T x ) { return x*x ; }

void main ()
{
  double [2] a = [ 1.0 , 2.0 ] ;

  writeln ( map ! ( sq ) ( a[] ) ) ;
}

I'm not sure if it's by design.

-- 
Dmitry Olshansky
August 21, 2010
Re: map on fixed-size arrays
On 08/21/2010 12:37 PM, Eduardo Cavazos wrote:
> Hello,
>
> The 'map' from std.algorithm doesn't seem to work with fixed-size arrays:
>
> ----------------------------------------------------------------------
> import std.stdio ;
> import std.math ;
> import std.algorithm ;
>
> T sq ( T ) ( T x ) { return x*x ; }
>
> void main ()
> {
> double [2] a = [ 1.0 , 2.0 ] ;
>
> writeln ( map ! ( sq ) ( a ) ) ;
> }
> ----------------------------------------------------------------------
>
> $ rdmd test_map_sq_fixed_size_b.d
> /usr/include/d/dmd/phobos/std/algorithm.d(108): Error: template instance
> Map!(sq,double[2u]) does not match template declaration Map(alias
> fun,Range) if (isInputRange!(Range))
>
> Is this an intended limitation?
>
> Ed

IIRC, it's intended. Use a[] to get a dynamic array from a.
August 21, 2010
Re: map on fixed-size arrays
On 8/21/10 5:37 CDT, Eduardo Cavazos wrote:
> Hello,
>
> The 'map' from std.algorithm doesn't seem to work with fixed-size arrays:
>
> ----------------------------------------------------------------------
> import std.stdio ;
> import std.math ;
> import std.algorithm ;
>
> T sq ( T ) ( T x ) { return x*x ; }
>
> void main ()
> {
> double [2] a = [ 1.0 , 2.0 ] ;
>
> writeln ( map ! ( sq ) ( a ) ) ;
> }
> ----------------------------------------------------------------------
>
> $ rdmd test_map_sq_fixed_size_b.d
> /usr/include/d/dmd/phobos/std/algorithm.d(108): Error: template instance
> Map!(sq,double[2u]) does not match template declaration Map(alias
> fun,Range) if (isInputRange!(Range))
>
> Is this an intended limitation?
>
> Ed

To some extent, yes; fixed-size arrays are passed by value and most of 
the time you don't want that with an algorithm. You have the burden to 
append "[]" to fixed-size arrays so they are passed inside the 
algorithms as dynamic-length slices.

std.algorithm could detect that and take care of that detail for you, at 
the cost of duplicating most function signatures.


Andrei
Top | Discussion index | About this forum | D home