www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - proposal: add bytesof Property, reduce pointer/cast requires

reply redsea <redsea 163.com> writes:
I found D's array and slice is very useful, most time I can use byte [] to
finish my work without using pointer.

But when I need to copy from/to a struct I need to use code like this:

byteary[i..i+structvar.sizeof] = (cast (byte *)&structvar)[0..structvar.sizeof)

the cast, the &, looking bad, and code is complex.

if d can supply a internal property, let's say bytes of, can give us the
correctly byte [] slice, then I can write:

byteary[i..i+struct.var.sizeof] = structvar.bytesof

and, when we write a function send struct to network, we can simple write:
void sendPacket(byte[] content);

and we can use as that:

sendPacket(structvar.bytesof);

any suggestions ?
Nov 17 2007
next sibling parent reply "Jarrett Billingsley" <kb3ctd2 yahoo.com> writes:
"redsea" <redsea 163.com> wrote in message 
news:fho4go$1r6t$1 digitalmars.com...

I found D's array and slice is very useful, most time I can use byte [] to 
finish my work without using pointer.

 But when I need to copy from/to a struct I need to use code like this:

 byteary[i..i+structvar.sizeof] = (cast (byte 
 *)&structvar)[0..structvar.sizeof)

 the cast, the &, looking bad, and code is complex.

 if d can supply a internal property, let's say bytes of, can give us the 
 correctly byte [] slice, then I can write:

 byteary[i..i+struct.var.sizeof] = structvar.bytesof

 and, when we write a function send struct to network, we can simple write:
 void sendPacket(byte[] content);

 and we can use as that:

 sendPacket(structvar.bytesof);

 any suggestions ?


byte[] bytesOf(T)(ref T t)
{
    return (cast(byte*)&t)[0 .. t.sizeof];
}

byteArray[i .. i + struct.var.sizeof] = bytesOf(structVar);
sendPacket(bytesOf(structVar));
Nov 17 2007
next sibling parent redsea <redsea 163.com> writes:
This help some, but it don't work at static array, and literal integer, read
following, the first call is ok, next 2 call is not ok, change ref to in cause
other problems, still not ok.  (d 1.022)

module test;

import tango.io.Stdout;
alias Stdout od;

byte[] bytesof(T)(ref T t)
{
    return (cast(byte*)&t)[0 .. t.sizeof];
}

struct SS
{
    int a;
    int b;
}

int main()
{
    SS [10] as;

    SS s;

    od(bytesof(s).length).newline;

    od(bytesof(as).length).newline;

    od(bytesof(3).length).newline;
    return 0;
}



Jarrett Billingsley Wrote:


 "redsea" <redsea 163.com> wrote in message 
 news:fho4go$1r6t$1 digitalmars.com...

I found D's array and slice is very useful, most time I can use byte [] to 
finish my work without using pointer.

 But when I need to copy from/to a struct I need to use code like this:

 byteary[i..i+structvar.sizeof] = (cast (byte 
 *)&structvar)[0..structvar.sizeof)

 the cast, the &, looking bad, and code is complex.

 if d can supply a internal property, let's say bytes of, can give us the 
 correctly byte [] slice, then I can write:

 byteary[i..i+struct.var.sizeof] = structvar.bytesof

 and, when we write a function send struct to network, we can simple write:
 void sendPacket(byte[] content);

 and we can use as that:

 sendPacket(structvar.bytesof);

 any suggestions ?


 
 byte[] bytesOf(T)(ref T t)
 {
     return (cast(byte*)&t)[0 .. t.sizeof];
 }
 
 byteArray[i .. i + struct.var.sizeof] = bytesOf(structVar);
 sendPacket(bytesOf(structVar));
Nov 17 2007
prev sibling next sibling parent "Janice Caron" <caron800 googlemail.com> writes:
 byte[] bytesOf(T)(ref T t)
 {
     return (cast(byte*)&t)[0 .. t.sizeof];
 }

 byteArray[i .. i + struct.var.sizeof] = bytesOf(structVar);
 sendPacket(bytesOf(structVar));


That's the way to do it!

But since bytesOf() operates on an array, one could presumably also write

    sendPacket(structVar.bytesOf);

which is what the orignal poster wanted. (Isn't D great?)
Nov 17 2007
prev sibling parent "Janice Caron" <caron800 googlemail.com> writes:
On 11/18/07, Janice Caron <caron800 googlemail.com> wrote:

 But since bytesOf() operates on an array, one could presumably also write

     sendPacket(structVar.bytesOf);

 which is what the orignal poster wanted. (Isn't D great?)


Nah. Bugger, I'm wrong. Sorry, retract that.
Anyway I still think Jarrett's idea is the perfect solution.
Nov 17 2007
prev sibling next sibling parent reply Robert Fraser <fraserofthenight gmail.com> writes:
redsea Wrote:


 if d can supply a internal property, let's say bytes of, can give us the
correctly byte [] slice


It's not a bad idea, but I can see it being used rarely enough it wouldn't be
necessary. Plus, if you're really serializing something for cross-platform use,
you should consider endianness, which neither your method nor bytesof would
capture.
Nov 17 2007
parent redsea <redsea 163.com> writes:
Robert Fraser Wrote:


 redsea Wrote:
 

 if d can supply a internal property, let's say bytes of, can give us the
correctly byte [] slice


 
 It's not a bad idea, but I can see it being used rarely enough it wouldn't be
necessary. Plus, if you're really serializing something for cross-platform use,
you should consider endianness, which neither your method nor bytesof would
capture.


I don't care endians in most case, because most of my application need high
performance, I don't want to waste CPU time when I use X86 servers (&clients). 
X86 serve is powerfull and cheap. 

At china, here are many mass online game providers, theire program would also
direct send struct for simplify and efficiency, don't care about endians too.

At, I  use share memory to do comunications between processes,  between a D
program and C++ program,  even send struct via netlink interface to linux
kernel, here we don't have endians  problem -- but the align problem want to be
check carefully, we write a  public .inc head file, include in c and 
mixin(import) in d, and a program to verify their size is match.

En, perhaps later, I would care about endians, because I maybe would use MIPS
network processor (hope gdc stable enough to write the management program on
the chip at that time), but there still more high bandwidth internal
communication don't need to care about endians.
Nov 17 2007
prev sibling next sibling parent reply "Lionello Lunesu" <lio lunesu.remove.com> writes:
The cast to void[] is implicit:

void[] bytesof = (&structvar)[0..1];

and there's no need for the .sizeof if you leave the pointer of type 
struct*.

It also makes sense to make the cast to byte[] NOT implicit, since the 
resulting array might not be portable on a system with different endianess.

L.
Nov 18 2007
parent redsea <redsea 163.com> writes:
This seems work with static arary,  the following code give result 36.

Not very simple, but typesafe.

Thanks,  everybody!
 

struct SS
{
    int i;
    int j;
    int k;
}


byte[]  tobytes(void [] v)
{
    return cast(byte []) v;
}

int main()
{
    SS[3] structvar;

    byte[] bytesof = tobytes((&structvar)[0..1]);
    Stdout(bytesof.length).newline;

    return 0;
}

Lionello Lunesu Wrote:


 The cast to void[] is implicit:
 
 void[] bytesof = (&structvar)[0..1];
 
 and there's no need for the .sizeof if you leave the pointer of type 
 struct*.
 
 It also makes sense to make the cast to byte[] NOT implicit, since the 
 resulting array might not be portable on a system with different endianess.
 
 L.
Nov 18 2007
prev sibling parent reply Henning Hasemann <hhasemann web.de> writes:
Maybe this helps you:

http://www.leetless.de/indiana-parts/s11n.d

I also have a more up-to-date version of that code available but
currently not uploaded, so mail me if you're interested in that one.

Henning

-- 
GPG Public Key:
http://gpg-keyserver.de/pks/lookup?op=get&search=0xDDD6D36D41911851
Nov 18 2007
parent redsea <redsea 163.com> writes:
Thanks very much, I download it, and maybe use somewhere else :)

Henning Hasemann Wrote:


 
 Maybe this helps you:
 
 http://www.leetless.de/indiana-parts/s11n.d
 
 I also have a more up-to-date version of that code available but
 currently not uploaded, so mail me if you're interested in that one.
 
 Henning
 
 -- 
 GPG Public Key:
 http://gpg-keyserver.de/pks/lookup?op=get&search=0xDDD6D36D41911851
Nov 19 2007