• bearophile (36/36) May 06 2010 While translating some old code I have found that allocating a fixed-siz...
• Lars T. Kyllingstad (6/8) May 06 2010 I don't know, but it was discussed not too long ago:
• bearophile (16/17) May 06 2010 Thank you, it seems Andrei agrees with me. But I think here thinks have ...
• Pelle (3/20) May 06 2010 Which is why they decided that the [] will always be needed for array

bearophile <bearophileHUGS lycos.com> writes:

While translating some old code I have found that allocating a fixed-size array
on the heap is not allowed, do you know why?


template Arr(int N) {
    alias int[N] Arr;
}
void main() {
    auto p = new Arr!(10);
}


The compiler says:
test.d(5): Error: new can only create structs, dynamic arrays or class objects,
not int[10u]'s

------------------

while this is OK, and allocates the same data on the heap:

struct Arr(int N) {
    int[N] data;
}
void main() {
    auto p = new Arr!(10);
}


Maybe it's a problem caused by conflicting syntax. This can be a way to avoid
the ambiguity:

auto p = new int[10];  => static array
auto p = new int[](10);  => dynamic array

------------------

I have also tried this, with not good results:


import std.stdio: writeln;
struct Arr(int N) {
    int[N] data;
    alias data this; // Bugs Bunny
}
void main() {
    auto p = new Arr!(10);
    writeln(*p);
}


toString(int[10u]) called from ...\dmd\src\phobos\std\conv.d(245) is
deprecated. Instead you may want to import std.conv and use to!string(x)
instead of toString(x).
toString(int[10u]) called from ...\dmd\src\phobos\std\conv.d(262) is
deprecated. Instead you may want to import std.conv and use to!string(x)
instead of toString(x).
Arr!(10)(0 0 0 0 0 0 0 0 0 0)

Bye and thank you,
bearophile

"Lars T. Kyllingstad" <public kyllingen.NOSPAMnet> writes:

On Thu, 06 May 2010 06:37:20 -0400, bearophile wrote:

 While translating some old code I have found that allocating a
 fixed-size array on the heap is not allowed, do you know why?

I don't know, but it was discussed not too long ago:

http://www.digitalmars.com/d/archives/digitalmars/D/
eliminate_new_operator_paraphernalia_106108.html

In particular, note Kasumi Hanazuki's post and Andrei's response to it.

-Lars

bearophile <bearophileHUGS lycos.com> writes:

Lars T. Kyllingstad:
 In particular, note Kasumi Hanazuki's post and Andrei's response to it.

Thank you, it seems Andrei agrees with me. But I think here thinks have to be
kept tidy, otherwise it's easy to make a mess.

The syntax offers various interesting possibilities for a future International
Obfuscated D Code Contest:

import std.stdio: writeln;
struct Arr(int N) {
    int[N] data;
    alias data this;
}
void main() {
    auto p = new Arr!(10);
    *p = 10;
    writeln(p.data); // Output: 10 10 10 10 10 10 10 10 10 10
}


This is not good :-(
Bye,
bearophile

Pelle <pelle.mansson gmail.com> writes:

On 05/06/2010 01:10 PM, bearophile wrote:
 Lars T. Kyllingstad:
 In particular, note Kasumi Hanazuki's post and Andrei's response to it.

 Thank you, it seems Andrei agrees with me. But I think here thinks have to be
kept tidy, otherwise it's easy to make a mess.

 The syntax offers various interesting possibilities for a future International
Obfuscated D Code Contest:

 import std.stdio: writeln;
 struct Arr(int N) {
      int[N] data;
      alias data this;
 }
 void main() {
      auto p = new Arr!(10);
      *p = 10;
      writeln(p.data); // Output: 10 10 10 10 10 10 10 10 10 10
 }


 This is not good :-(
 Bye,
 bearophile

Which is why they decided that the [] will always be needed for array 
operations. Right?