Hi,

Currently, Phobos has Tuple but doesn't have tie.
So I implemented tie for std.typecons.Tuple.

http://www.bitbucket.org/repeatedly/scrap/src/tip/tie_.d

Original author is 9rnsr but his version is a bit messy and overkill.
I substantially rewrote his tie.
- clear syntax (tie(&n, &d) => tie(n, d))
- Implicit conversion

I would like to suggest this tie is integrated into phobos.

What do you think of this?

On Mon, May 10, 2010 at 2:22 PM, Masahiro Nakagawa <repeatedly at gmail.com> wrote:
> Hi,
>
> Currently, Phobos has Tuple but doesn't have tie.
> So I implemented tie for std.typecons.Tuple.
>
> http://www.bitbucket.org/repeatedly/scrap/src/tip/tie_.d
>
> Original author is 9rnsr but his version is a bit messy and overkill.
> I substantially rewrote his tie.
> - clear syntax (tie(&n, &d) => tie(n, d))
> - Implicit conversion
>
> I would like to suggest this tie is integrated into phobos.
>
> What do you think of this?
> _______________________________________________
> phobos mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos
>

Looks great. Just 2 questions:
1) Does chaining works? I.e.
tie(a, b) = tie(c, d) = tuple(1,2); // a = b = 1, b = c = 2

2) You could probably use __dollar for ignored elements, to prevent
namespace pollution and symbol conflicts:
tie(a, $) = tuple(1, 2); // a = 1

This leads to another idea: Phobos should have struct Dolllar and global __dollar variable so that everyone don't have to redefine it, just import some Phobos module and use it.

On Mon, 10 May 2010 21:32:00 +0900, Denis <2korden at gmail.com> wrote:

> On Mon, May 10, 2010 at 2:22 PM, Masahiro Nakagawa <repeatedly at gmail.com> wrote:
>> Hi,
>>
>> Currently, Phobos has Tuple but doesn't have tie.
>> So I implemented tie for std.typecons.Tuple.
>>
>> http://www.bitbucket.org/repeatedly/scrap/src/tip/tie_.d
>>
>> Original author is 9rnsr but his version is a bit messy and overkill.
>> I substantially rewrote his tie.
>> - clear syntax (tie(&n, &d) => tie(n, d))
>> - Implicit conversion
>>
>> I would like to suggest this tie is integrated into phobos.
>>
>> What do you think of this?
>>
>
> Looks great. Just 2 questions:
> 1) Does chaining works? I.e.
> tie(a, b) = tie(c, d) = tuple(1,2); // a = b = 1, b = c = 2

http://www.bitbucket.org/repeatedly/scrap/changeset/e4c270e6c30d Done.

> 2) You could probably use __dollar for ignored elements, to prevent
> namespace pollution and symbol conflicts:
> tie(a, $) = tuple(1, 2); // a = 1

I don't understand this. Please show me example.
I tried following code with dmd 2.046 beta:
-----
import std.stdio;

struct Dollar {}
__gshared Dollar __dollar;

void main()
{
     writeln($);
}

// foo.d(8): '$' is valid only inside [] of index or slice
-----
Umm...

> This leads to another idea: Phobos should have struct Dolllar and global __dollar variable so that everyone don't have to redefine it, just import some Phobos module and use it.

I agree.

On Mon, May 10, 2010 at 7:44 PM, Masahiro Nakagawa <repeatedly at gmail.com> wrote:
> On Mon, 10 May 2010 21:32:00 +0900, Denis <2korden at gmail.com> wrote:
>
>> 2) You could probably use __dollar for ignored elements, to prevent
>> namespace pollution and symbol conflicts:
>> tie(a, $) = tuple(1, 2); // a = 1
>
> I don't understand this. Please show me example.
> I tried following code with dmd 2.046 beta:
> -----
> import std.stdio;
>
> struct Dollar {}
> __gshared Dollar __dollar;
>
> void main()
> {
> ? ?writeln($);
> }
>
> // foo.d(8): '$' is valid only inside [] of index or slice
> -----
> Umm...
>

My bad, it is indeed disallowed outside of opSlice. Didn't know that.

I think tie() should be removed because it is unsafe. Masahiro, could you please look into that?

We should discuss alternative means, but I strongly believe whatever abstractions we add should be, if at all possible, safe. But tie() is not appropriate.


Andrei

On 05/10/2010 05:22 AM, Masahiro Nakagawa wrote:
> Hi,
>
> Currently, Phobos has Tuple but doesn't have tie.
> So I implemented tie for std.typecons.Tuple.
>
> http://www.bitbucket.org/repeatedly/scrap/src/tip/tie_.d
>
> Original author is 9rnsr but his version is a bit messy and overkill.
> I substantially rewrote his tie.
> - clear syntax (tie(&n, &d) => tie(n, d))
> - Implicit conversion
>
> I would like to suggest this tie is integrated into phobos.
>
> What do you think of this?
> _______________________________________________
> phobos mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos