phobos commit, revision 2018


user: rsinfu

msg:
std.typecons: Several improvements to Tuple.

- Tuple members are now accessible with the syntax a[0], a[1] etc.
- Eliminated an internal union (see bug 4421 and 4846). Re-implemented named members with alias.
- Worked around bug 4424. Got opAssign back.
- Made Tuple.slice!(from, to) to preserve field names if any.
- Added isTuple!(T) template.
- Removed the 'default constructor' that is never called.
- Changed Tuple.toString() to use formattedWrite() instead of the `static if (is(to!string...))` thingy.

http://www.dsource.org/projects/phobos/changeset/2018

Apparently the bug 2800 have been fixed at some point, so I enabled the alias this in the Tuple.  Does this changeset cause any regression bug? In my local Phobos repo, I changed all x.field[k] in the modules to x[k] and they seemed to work fine.

( http://d.puremagic.com/issues/show_bug.cgi?id=2800 )


Shin

Yes yes yes! I am so happy bug 2800 was fixed. This is a huge step towards legitimizing Tuple. I cringed whenever I had to write t.field[0] instead of the natural t[0].

BTW: a large part of the motivation for "alias this" was allowing this very idiom. True story!


Andrei

On 09/17/2010 04:40 PM, dsource.org wrote:
> phobos commit, revision 2018
>
>
> user: rsinfu
>
> msg:
> std.typecons: Several improvements to Tuple.
>
> - Tuple members are now accessible with the syntax a[0], a[1] etc.
> - Eliminated an internal union (see bug 4421 and 4846). Re-implemented named members with alias.
> - Worked around bug 4424. Got opAssign back.
> - Made Tuple.slice!(from, to) to preserve field names if any.
> - Added isTuple!(T) template.
> - Removed the 'default constructor' that is never called.
> - Changed Tuple.toString() to use formattedWrite() instead of the `static if (is(to!string...))` thingy.
>
> http://www.dsource.org/projects/phobos/changeset/2018
>
> _______________________________________________
> phobos mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos

Andrei Alexandrescu <andrei at erdani.com> wrote:

> Yes yes yes! I am so happy bug 2800 was fixed. This is a huge step towards legitimizing Tuple. I cringed whenever I had to write t.field[0] instead of the natural t[0].
>
> BTW: a large part of the motivation for "alias this" was allowing this very idiom. True story!

I don't really post here, but this is too awesome. Finally!

-- 
Simen

On Sat, Sep 18, 2010 at 00:58, Simen Kjaeraas <simen.kjaras at gmail.com>wrote:

> Andrei Alexandrescu <andrei at erdani.com> wrote:
>
>  Yes yes yes! I am so happy bug 2800 was fixed. This is a huge step towards
>> legitimizing Tuple. I cringed whenever I had to write t.field[0] instead of the natural t[0].
>>
>> BTW: a large part of the motivation for "alias this" was allowing this very idiom. True story!
>>
>
> I don't really post here, but this is too awesome. Finally!
>


Yeaaaahhhhh!!!
Oh, that's good news indeed. (and an errata less for TDPL). I just tested a
Tuple-equivalent in DMD 2.049 and it works OK.

Oh, and I see that

auto t = tuple();

now works. That's good! It's a common degenerate case with variadic
templates, and Tuple!() is a good unit type (ie: a type with only one value,
namely tuple() / Tuple!()() )


I just tested alias someTuple this with a function call and it doesn't work. Oh well, that's for another time.

int foo(int i, double d, string s) { return 0;}
auto t = tuple(1, 3.1415, "abc");
auto r = foo(t); // doesn't work.



Philippe
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puremagic.com/pipermail/phobos/attachments/20100918/4bc0a0b2/attachment.html>