220 16645 article Path: news.gmane.org!not-for-mail From: Zijie He Newsgroups: gmane.comp.lang.c++.isocpp.proposals Subject: Re: Propose a smart convertor function Date: Tue, 3 Mar 2015 06:58:16 -0800 (PST) Lines: 175 Approved: news@gmane.org Message-ID: References: <1e329e3e-1c24-4df7-a8a4-8a196e0bc5aa@isocpp.org> <42516db7-6c38-43c5-b2e7-f414e6b78139@isocpp.org> <79186eb1-219a-46ff-9586-c4a373c3b172@isocpp.org> Reply-To: std-proposals@isocpp.org NNTP-Posting-Host: plane.gmane.org Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="----=_Part_70_2024141473.1425394696839" X-Trace: ger.gmane.org 1425394707 5650 80.91.229.3 (3 Mar 2015 14:58:27 GMT) X-Complaints-To: usenet@ger.gmane.org NNTP-Posting-Date: Tue, 3 Mar 2015 14:58:27 +0000 (UTC) To: std-proposals@isocpp.org Original-X-From: std-proposals+bncBD6IXSMN7QKRBCMY26TQKGQE32FUJHA@isocpp.org Tue Mar 03 15:58:26 2015 Return-path: Envelope-to: gclcip-std-proposals@m.gmane.org Original-Received: from mail-yk0-f200.google.com ([209.85.160.200]) by plane.gmane.org with esmtp (Exim 4.69) (envelope-from ) id 1YSoHB-0005He-U5 for gclcip-std-proposals@m.gmane.org; Tue, 03 Mar 2015 15:58:26 +0100 Original-Received: by ykbq200 with SMTP id q200sf121126472ykb.3 for ; Tue, 03 Mar 2015 06:58:18 -0800 (PST) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20130820; h=x-gm-message-state:date:from:to:message-id:in-reply-to:references :subject:mime-version:content-type:x-original-sender:reply-to :precedence:mailing-list:list-id:list-post:list-help:list-archive :list-subscribe:list-unsubscribe; bh=p7e31KHEXrqtrybRcNaWtqtC7B1HZ/CObW0HoShEzAY=; b=gxE/Nrp4TC04Br/7joZulkynLV9Lbu4q7pKcfqjAcXwJHq9uUO1o+61Oe/FOi4NoM2 0b8TN3BV93ceeStIgxhsZoIxhiB5Vm8KujjraW9c6qH37Qb76YCV4pBkLU9oiEvpQTzl aX66sE29e40GOGV/r3CLmKmQV69fkqZCijmubwTvaIloOeQWJVVlZGuL8XQ7JZUbdl+u TbfS2SMxeTMmSbdsFuPM6YT1u+2hkRkIyQ2pJUOCIyb9TBAerpyxxSk0u54Rywqoh8lu XVuMtrUeOAPxPyaeCzQLJDnun0A4jz6jf76ODvrK01nSBl69zNy5D/F1pVO1yggEnfxX BXNw== X-Gm-Message-State: ALoCoQlzHCcSsLLPaUhCm3+HXJ5i2ayHyrJWlAB2P9253yXYH3O21y8pE6f8PHSMe9ocIq2kHdga X-Received: by 10.140.196.209 with SMTP id r200mr31729526qha.0.1425394698026; Tue, 03 Mar 2015 06:58:18 -0800 (PST) X-BeenThere: std-proposals@isocpp.org Original-Received: by 10.182.149.164 with SMTP id ub4ls40660obb.98.gmail; Tue, 03 Mar 2015 06:58:17 -0800 (PST) X-Received: by 10.182.65.169 with SMTP id y9mr226733obs.17.1425394697264; Tue, 03 Mar 2015 06:58:17 -0800 (PST) In-Reply-To: <79186eb1-219a-46ff-9586-c4a373c3b172@isocpp.org> X-Original-Sender: Hzj_jie@hotmail.com Precedence: list Mailing-list: list std-proposals@isocpp.org; contact std-proposals+owners@isocpp.org List-ID: X-Google-Group-Id: 399137483710 List-Post: , List-Help: , List-Archive: List-Subscribe: , List-Unsubscribe: , Xref: news.gmane.org gmane.comp.lang.c++.isocpp.proposals:16645 Archived-At: ------=_Part_70_2024141473.1425394696839 Content-Type: multipart/alternative; boundary="----=_Part_71_1457284584.1425394696839" ------=_Part_71_1457284584.1425394696839 Content-Type: text/plain; charset=UTF-8 Sorry, do you really mean c2 = 1; does not use operator=(int)? If there is a conversion operator, it will be used, but in the case I have provided, the operator= is definitely involved. And I still cannot quite understand, why you think implicit constructor and operator= are the same thing. They are two different functions, while C c = 1; uses implicit constructor, C c; c = 1; uses operator=. A class may implement one of them or both, or none. Some school books mentioned that you always should implement operator= (const T&) with T(const T&), but the compiler does not force you to do so. On Tuesday, March 3, 2015 at 2:13:07 PM UTC+8, Nicol Bolas wrote: > > > > On Monday, March 2, 2015 at 9:28:25 PM UTC-5, Zijie He wrote: >> >> The proposal is targeting three different problems, >> 1. is_assignable is much more than T1 t1 = t2; >> > > Actually, that's line is not an example of is_assignable at all. That > statement will never use the assignment operator; it will instead use an > appropriate conversion (either via an implicit constructor or via a single > implicit conversion operator call). > > >> 2. As you have mentioned, performance concern. There may be several ways >> to convert from T1 to T2, while this implementation can help to select the >> best one. >> > > As previously stated, the only performance concern as things currently > stand is the rare case of performing a conversion into a live object. And > you have yet to explain why this case is important enough to bother with > such a change. > > >> 3. Typically in template programming, usually you do not know the types >> of T1 and T2, while there may be several ways to convert from T1 to T2. >> > > There are not "several" ways to convert; there are precisely two: either > via initialization of a new value or via copy/move-into-a-live-object. Both > ways in C++ use the assignment operator (though the initialization case > doesn't call the overload for it). The language supports these cases well > enough without template metaprogramming or 'convert_to' or other such stuff. > > So what are these "several ways" you keep talking about optimizing for? > What are the other alternatives? > > Several types may implements constructor, several types may implements >> operator=. The behavior will be same, but in the template, you may need to >> handle several cases. So this implementation is to separate the irrelevant >> logic from this kind of templates. >> > > Nonsense. In a template, you impose some particular requirement on the > type. Sans-concepts, this requirement is implicit. If you template needs to > "convert" from one type to another, we have an acceptable syntax to do that: > > T1 t1 = t2; > > There are no performance issues with this. There is no "irrelevant logic" > needed to make this work. You don't need any special cases. You simply use > the language features. > > The user will implement the implicit conversion in T1 or T2 at their > leisure, and your template code will adapt to their choice. There are no > cases for you to consider. So long as the user has done their jobs in > making T1 and T2, this will work. > > And if they didn't, then they get a compiler error. As they should. > > If you want to be nicer to the user (so the conversions don't have to be > implicit), you invoke direct initialization: > > T1 t1(t2); > -- --- You received this message because you are subscribed to the Google Groups "ISO C++ Standard - Future Proposals" group. To unsubscribe from this group and stop receiving emails from it, send an email to std-proposals+unsubscribe@isocpp.org. To post to this group, send email to std-proposals@isocpp.org. Visit this group at http://groups.google.com/a/isocpp.org/group/std-proposals/. ------=_Part_71_1457284584.1425394696839 Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable
Sorry, do you really mean c2 =3D 1; does not use operator= =3D(int)? If there is a conversion operator, it will be used, but in the ca= se I have provided, the operator=3D is definitely involved.
And I still= cannot quite understand, why you think implicit constructor and operator= =3D are the same thing. They are two different functions, while C c =3D 1; = uses implicit constructor, C c; c =3D 1; uses operator=3D. A class may impl= ement one of them or both, or none. Some school books mentioned that you al= ways should implement operator=3D (const T&) with T(const T&), but = the compiler does not force you to do so.

On Tuesday, March 3, 201= 5 at 2:13:07 PM UTC+8, Nicol Bolas wrote:


On Monday, March 2, 2015 at 9:28:25 PM = UTC-5, Zijie He wrote:
The proposal is targeting three different problems,
1. is_assignabl= e is much more than T1 t1 =3D t2;

Actually= , that's line is not an example of is_assignable at all. That statement wil= l never use the assignment operator; it will instead use an appropriate conversion (ei= ther via an implicit constructor or via a single implicit conversion operat= or call).
 
2. As you have mentioned, performance concern. There may be severa= l ways to convert from T1 to T2, while this implementation can help to sele= ct the best one.

As previously stated, the o= nly performance concern as things currently stand is the rare case of perfo= rming a conversion into a live object. And you have yet to explain why this= case is important enough to bother with such a change.
 
3. Typically in t= emplate programming, usually you do not know the types of T1 and T2, while = there may be several ways to convert from T1 to T2.

There are not "several" ways to convert; there are precisely two:= either via initialization of a new value or via copy/move-into-a-live-obje= ct. Both ways in C++ use the assignment operator (though the initialization= case doesn't call the overload for it). The language supports these cases = well enough without template metaprogramming or 'convert_to' or other such = stuff.

So what are these "several ways" you keep talking about optim= izing for? What are the other alternatives?

Several types may implements c= onstructor, several types may implements operator=3D. The behavior will be = same, but in the template, you may need to handle several cases. So this im= plementation is to separate the irrelevant logic from this kind of template= s.

Nonsense. In a template, you impose= some particular requirement on the type. Sans-concepts, this requirement i= s implicit. If you template needs to "convert" from one type to another, we= have an acceptable syntax to do that:

T1 t1 =3D t2;

There ar= e no performance issues with this. There is no "irrelevant logic" needed to= make this work. You don't need any special cases. You simply use the langu= age features.

The user will implement the implicit conversion in T1 = or T2 at their leisure, and your template code will adapt to their choice. = There are no cases for you to consider. So long as the user has done their = jobs in making T1 and T2, this will work.

And if they didn't, then t= hey get a compiler error. As they should.

If you want to be nicer to= the user (so the conversions don't have to be implicit), you invoke direct= initialization:

T1 t1(t2);

--

---
You received this message because you are subscribed to the Google Groups &= quot;ISO C++ Standard - Future Proposals" group.
To unsubscribe from this group and stop receiving emails from it, send an e= mail to std-proposa= ls+unsubscribe@isocpp.org.
To post to this group, send email to std-proposals@isocpp.org.
Visit this group at http://groups.google.com/a/isocpp.org/group/std-proposals/<= /a>.
------=_Part_71_1457284584.1425394696839-- ------=_Part_70_2024141473.1425394696839-- .