220 19872 <55D2BEC3.9070307@eudoxos.de> article Path: news.gmane.org!not-for-mail From: Roland Bock Newsgroups: gmane.comp.lang.c++.isocpp.proposals Subject: Re: Re: default arguments Date: Tue, 18 Aug 2015 07:12:35 +0200 Lines: 597 Approved: news@gmane.org Message-ID: <55D2BEC3.9070307@eudoxos.de> References: <736eb378-52a4-43ba-b3b5-a73719ba70af@isocpp.org> <86717349-1fce-4247-9372-08eedce4cd30@isocpp.org> Reply-To: std-proposals@isocpp.org NNTP-Posting-Host: plane.gmane.org Mime-Version: 1.0 Content-Type: multipart/alternative; boundary="------------010403090403020403070202" X-Trace: ger.gmane.org 1439874767 24071 80.91.229.3 (18 Aug 2015 05:12:47 GMT) X-Complaints-To: usenet@ger.gmane.org NNTP-Posting-Date: Tue, 18 Aug 2015 05:12:47 +0000 (UTC) To: std-proposals@isocpp.org Original-X-From: std-proposals+bncBCO7PYWVVELRBRP5ZKXAKGQE75N6OVI@isocpp.org Tue Aug 18 07:12:39 2015 Return-path: Envelope-to: gclcip-std-proposals@m.gmane.org Original-Received: from mail-lb0-f199.google.com ([209.85.217.199]) by plane.gmane.org with esmtp (Exim 4.69) (envelope-from ) id 1ZRZCQ-0003KO-O5 for gclcip-std-proposals@m.gmane.org; Tue, 18 Aug 2015 07:12:38 +0200 Original-Received: by lbcli1 with SMTP id li1sf51866352lbc.2 for ; Mon, 17 Aug 2015 22:12:38 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20130820; h=x-gm-message-state:message-id:date:from:user-agent:mime-version:to :subject:references:in-reply-to:content-type:x-original-sender :x-original-authentication-results:reply-to:precedence:mailing-list :list-id:x-spam-checked-in-group:list-post:list-help:list-archive :list-subscribe:list-unsubscribe; bh=WUikQ8oXJA1e2QyGIoNMKelX6OGPOqeIGd8rx1V2vX0=; b=NzX21Ry4XBG7gZhrzLBw+gd8h9Ra/JhBKbIsMbi04Zlim+mnjkIiaX6KigmyEC9/TY i7VXaAQlM22pm68MDHpTDkCK6wVovQ/oxxtOqwoH7+tgscyJnD7N5RR0WB76RBMTVQlL RP22XfkTzvCONId3munPKOU+Is+tu4i/T2MxMbpS7JG4IWKKf7FUV38owmFz2WZJml3D igrX0/d+LMyLb20IlS5aNy/T2mHx1CIAivCgddxUoDJOlnRTBwXPTmq+yfurNcMjO4Kv Btp94zDWf6JrrVSrKebNpZxntH8AZBgjOqpioMW249Ro3QUZcHcC89KGgLccz5Jci7mF HKAw= X-Gm-Message-State: ALoCoQneJnWlwtAds5LjF2WEspINTcO+zV61/FlNI7sd17Y0f1/PApev9rGLhBdwwZWhvTG7TeQq X-Received: by 10.194.93.198 with SMTP id cw6mr1238871wjb.3.1439874758264; Mon, 17 Aug 2015 22:12:38 -0700 (PDT) X-BeenThere: std-proposals@isocpp.org Original-Received: by 10.180.90.209 with SMTP id by17ls708892wib.12.canary; Mon, 17 Aug 2015 22:12:37 -0700 (PDT) X-Received: by 10.194.117.5 with SMTP id ka5mr9140497wjb.50.1439874757124; Mon, 17 Aug 2015 22:12:37 -0700 (PDT) Original-Received: from mout.kundenserver.de (mout.kundenserver.de. [212.227.126.187]) by mx.google.com with ESMTPS id eu3si24836567wic.47.2015.08.17.22.12.37 for (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Mon, 17 Aug 2015 22:12:37 -0700 (PDT) Received-SPF: neutral (google.com: 212.227.126.187 is neither permitted nor denied by best guess record for domain of rbock@eudoxos.de) client-ip=212.227.126.187; Original-Received: from [172.30.200.118] ([194.97.158.70]) by mrelayeu.kundenserver.de (mreue004) with ESMTPSA (Nemesis) id 0Ll1qT-1YtbQP0ZfB-00al4t for ; Tue, 18 Aug 2015 07:12:36 +0200 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.8.0 In-Reply-To: <86717349-1fce-4247-9372-08eedce4cd30@isocpp.org> X-Provags-ID: V03:K0:RUKkBehIuPtDt8voFKPyDBYy0QF0NY+Zz2Qb0lTDA0gObf++b5u WxmrJLFgXevzKgahXCDQYgKrG9uQZbsbWtl98OvZ1POsFpES++jGUI2EjFHYvNQlWaKuy9y cBcAU6Wrk2EdJtfSz1sUaq6ydSBnza9p84U5rwTCf0Dm1Tmrm2hG8IBNeQaJQjGOR/ABJ5M S+ILsG71pRp9s991YK57Q== X-UI-Out-Filterresults: notjunk:1;V01:K0:y90/5qpZCTc=:y+7urK71LVZQBJoL216lj8 tAxQDpE3ndZ3AiYs1PArfiWyJtgOKeEh0R3QeAWX5R72YsJR+P9L1Jzqv/Qq+O7OvOdyQaNaL yCvlag3hUOnRIJQoMrTgAIqxR+R4CsYE5YdIC/RuOTUnkTJ9VYzRR/iDDuLQsKVLIkSVHlv0s HZg6Gi4cCs1wndP04xzOivhv1Ypp2Hb7z8X3IPHgSUT2oGhl9Xr/66L2xppU9ybk+/0b+4j1j gLwbos2/W5VAazVP5D08cDRnUINv5jtyHs6zDSUufVjfSmLH66lT8A/HvT0wHjDI/lI3lKrBL WgCLxOBHqr2p7fu0iQq1t2JgEY/9v4izjcM8imbzSKw/yn/DRNAVrHcUoxvIA6TKw2wOj7GBf zDjeCxdmBDtt2VOjt2rqdPYdQ94v3I9cpw8Xzk/YznrB+5L1BdcEb9BdmYZ9zDJPhzyXvSt5y RfhFn/OXwwrV+tEhGaByNip3+Hjqnz7N6hr6fN6UFSg1Md+rHgVfaFcrd95jZT9JaCChiWfMU nV+zenWeCu1UNBIxQR+lHBWS3gJ2iB2XH6IRHg7/t972Q48bKSt240x8Mg4bfD0jsN+jQ+7tE wgca4RA980YfDCIkMtkqQo+apHHYRNIIxyxBN75zCUD4Gd1jNJbybAnxBWUZgV23CI3iwGEnL m3CVajVffuhk4PW04IhzXDtI/837FTA1f+zrcEBI6t4/s3ItrH01ON+vFHCWVj7ZSnyVvxYg8 O5nrUMc11VnTqB0g X-Original-Sender: rbock@eudoxos.de X-Original-Authentication-Results: mx.google.com; spf=neutral (google.com: 212.227.126.187 is neither permitted nor denied by best guess record for domain of rbock@eudoxos.de) smtp.mailfrom=rbock@eudoxos.de Precedence: list Mailing-list: list std-proposals@isocpp.org; contact std-proposals+owners@isocpp.org List-ID: X-Spam-Checked-In-Group: std-proposals@isocpp.org 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:19872 Archived-At: This is a multi-part message in MIME format. --------------010403090403020403070202 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable On 2015-08-18 00:52, Arthur O'Dwyer wrote: > I work on a codebase where this feature would have lots of use-cases, > because we have lots of classes whose constructors have defaulted > parameters: > > C(int buffer_size =3D 1024, int timeout =3D 10, int tries =3D 3) : > buffer_size_(buffer_size), timeout_(timeout), tries_(tries) {} > > auto cptr =3D new C(); // usually the defaults are okay > auto c2ptr =3D new C(2048); // use a bigger buffer > auto c3ptr =3D new C(1024, 1); // use the default buffer size but > with a shorter timeout > auto c4ptr =3D new C(2048, 10, 5); // testing: bump up the number > of tries > > It would be marginally more convenient (and refactor-proof) to be able > to write > > auto cptr =3D new C(); // usually the defaults are okay > auto c2ptr =3D new C(2048); // use a bigger buffer > auto c3ptr =3D new C(default, 1); // use the default buffer size > but with a shorter timeout > auto c4ptr =3D new C(2048, default, 5); // testing: bump up the > number of tries > > However, I consider defaulted parameters in general to be a huge > antipattern, and this constructor pattern in particular to be even worse. +10 (at least) > I would MUCH rather just rewrite all these constructors so that they > could be used as > > auto cptr =3D new C(); // usually the defaults are okay > auto c2ptr =3D new C().withBufferSize(2048); // use a bigger buffer > auto c3ptr =3D new C().withTimeout(1); // use the default buffer > size but with a shorter timeout > auto c4ptr =3D new C().withBufferSize(2048).withTries(5); // > testing: bump up the number of tries > > or > > auto cptr =3D new C(); // usually the defaults are okay > auto c2ptr =3D new C(C::BufferSize(2048)); // use a bigger buffer > auto c3ptr =3D new C(C::Timeout(1)); // use the default buffer size > but with a shorter timeout > auto c4ptr =3D new C(C::BufferSize(2048), C::Tries(5)); // testing: > bump up the number of tries Or (if you are definining individual types for your parameters anyway), why not auto cptr =3D new C(); // usually the defaults are okay auto c2ptr =3D new C(bufferSize =3D 2048); // use a bigger buffer auto c3ptr =3D new C(timeout =3D 1); // use the default buffer size bu= t with a shorter timeout auto c4ptr =3D new C(bufferSize =3D 2048, tries =3D 5); // testing: bu= mp up the number of tries You would have but one constructor (assuming Concepts Lite and Folding Expressions): template void operator()(T& t) { t.buffer_size_ =3D value;// or t.setBufferSize(value) } }; struct BufferSize { BufferSizeAssignment operator=3D(int value) { return {value}; } }; constexpr auto bufferSize =3D BufferSize{}; Of course, this would be much more convenient if we could configure names in template parameters and could write constexpr auto bufferSize =3D std::named_parameter; constexpr auto timeout =3D std::named_parameter; constexpr auto tries =3D std::named_parameter; Still trying to find the time to write the proposal for that... Best, Roland > > or (the most conservative way to rewrite our existing code) > > static constexpr int default_buffer_size =3D 1024; > static constexpr int default_timeout =3D 10; > static constexpr int default_tries =3D 3; > C() : buffer_size_(default_buffer_size), > timeout_(default_timeout), tries_(default_tries) {} > C(int buffer_size, int timeout, int tries) : > buffer_size_(buffer_size), timeout_(timeout), tries_(tries) {} > > auto cptr =3D new C(); // usually the defaults are okay > auto c2ptr =3D new C(2048, C::default_timeout, C::default_tries); > // use a bigger buffer > auto c3ptr =3D new C(C::default_buffer_size, 1, C::default_tries); > // use the default buffer size but with a shorter timeout > auto c4ptr =3D new C(2048, C::default_timeout, 5); // testing: bump > up the number of tries > > I don't really want to make it /easier/ for people to write the > "function taking tons of defaulted parameters" antipattern. > Let's teach people how to avoid the use of defaulted parameters, instead. > > =E2=80=93Arthur > > > On Monday, August 17, 2015 at 11:58:41 AM UTC-7, Bo Persson wrote: > > On 2015-08-17 20:37, Arthur Tchaikovsky wrote: > > I would really appreciate some feedback from people, or is this > proposal > > so uninteresting that there is no point in wasting keyboard on it? > > You might wonder how widely used it will be? > > If not all parameters are of the same type, you can already get > some of > the effect with a few overloads: > > void f(int a =3D 0, string b =3D "1", > complex c =3D 2.0, type4 d =3D 3); > > inline void f(int a, complex c) > { f(a, "1", c, 3); } > > inline void f(string b, complex c) > { f(0, b, c, 3); } > > assuming that d is put last because you hardly ever want to change > the > default. > > Just saying that you can already do this without a language change. > > > Bo Persson > > > > > > > On Saturday, 15 August 2015 17:49:50 UTC+1, Arthur Tchaikovsky > wrote: > > > > Hi, > > Does anyone see helpfulness of such solution: > > Having fnc: > > > > void f(int a =3D 0, int b =3D 1, int c =3D 2, int d =3D 3); > > > > I think that it would be nice to be able to say whilst > calling it > > and requiring only some of those args to be non-default: > > > > f(default,2,default,4); > > > > Thoughts? > > > > > > > --=20 > > --- > 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/. --=20 ---=20 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 e= mail 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-proposa= ls/. --------------010403090403020403070202 Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable
On 2015-08-18 00:52, Arthur O'Dwyer wrote:
I work on a codebase where this feature would have lots of use-cases, because we have lots of classes whose constructors have defaulted parameters:

=C2=A0 =C2=A0 C(int buffer_size =3D 1024, int timeout =3D 10, = int tries =3D 3) : buffer_size_(buffer_size), timeout_(timeout), tries_(tries) {}

=C2=A0 =C2=A0 auto cptr =3D new C(); =C2=A0// usually the defa= ults are okay
=C2=A0 =C2=A0 auto c2ptr =3D new C(2048); =C2=A0// use a bigge= r buffer
=C2=A0 =C2=A0 auto c3ptr =3D new C(1024, 1); =C2=A0// use the def= ault buffer size but with a shorter timeout
=C2=A0 =C2=A0 auto c4ptr =3D new C(2048, 10, 5); =C2=A0// test= ing: bump up the number of tries

It would be marginally more convenient (and refactor-proof) to be able to write

=C2=A0 =C2=A0 auto cptr =3D new C(); =C2=A0// usually the de= faults are okay
=C2=A0 =C2=A0 auto c2ptr =3D new C(2048); =C2=A0// use a big= ger buffer
=C2=A0 =C2=A0 auto c3ptr =3D new C(default, 1); =C2=A0// use th= e default buffer size but with a shorter timeout
=C2=A0 =C2=A0 auto c4ptr =3D new C(2048,=C2=A0default, 5); = =C2=A0// testing: bump up the number of tries

However, I consider defaulted parameters in general to be a huge antipattern, and this constructor pattern in particular to be even worse.

+10 (at least)

I would MUCH rather just rewrite all these constructors so that they could be used as

=C2=A0 =C2=A0 auto cptr =3D new C(); =C2=A0// usually the de= faults are okay
=C2=A0 =C2=A0 auto c2ptr =3D new C().withBufferSize(2048); = =C2=A0// use a bigger buffer
=C2=A0 =C2=A0 auto c3ptr =3D new C().withTimeout(1); =C2=A0// u= se the default buffer size but with a shorter timeout
=C2=A0 =C2=A0 auto c4ptr =3D new C().withBufferSize(2048).withTries(5); =C2=A0// testing: bump u= p the number of tries

or

=C2=A0 =C2=A0 auto cptr =3D new C(); =C2=A0// usually the = defaults are okay
=C2=A0 =C2=A0 auto c2ptr =3D new C(C::BufferSize(2048)); = =C2=A0// use a bigger buffer
=C2=A0 =C2=A0 auto c3ptr =3D new C(C::Timeout(1)); =C2=A0// u= se the default buffer size but with a shorter timeout
=C2=A0 =C2=A0 auto c4ptr =3D new C(C::BufferSize(2048), C::Tries(5)); =C2=A0// testing: bump up the number of tries

Or (if you are definining individual types for your parameters anyway), why not

=C2=A0 =C2=A0 auto cptr =3D new C(); =C2=A0// usually the de= faults are okay
=C2=A0 =C2=A0 auto c2ptr =3D new C(bufferSize =3D 2048); =C2= =A0// use a bigger buffer
=C2=A0 =C2=A0 auto c3ptr =3D new C(timeout =3D 1); =C2=A0/= / use the default buffer size but with a shorter timeout
=C2=A0 =C2=A0 auto c4ptr =3D new C(bufferSize =3D 2048, trie= s =3D 5); =C2=A0// testing: bump up the number of tries

You would have but one constructor (assuming Concepts Lite and Folding Expressions):

template<Assignment... A)
C(const A&... a) : buffer_size_(1024), timeout_(10), tries_(3)
{
=C2=A0 (a(*this),...);
}

And some types/constexprs:

struct BufferSizeAssignment
{
=C2=A0=C2=A0 int value;

=C2=A0=C2=A0 template<typename T>
=C2=A0=C2=A0 void operator()(T& t)
=C2=A0=C2=A0 {
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 t.buffer_size_ =3D value;<= tt> // or t.setBufferSize(value)
=C2=A0=C2=A0 }
};

struct BufferSize
{
=C2=A0=C2=A0 BufferSizeAssignment operator=3D(int value)<= tt>
=C2=A0=C2=A0 {
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 return {value};
=C2=A0=C2=A0 }
};

constexpr auto bufferSize =3D BufferSize{};


Of course, this would be much more convenient if we could configure names in template parameters and could write

constexpr auto bufferSize =3D std::named_parameter<int, `buffer_size_`>;
 constexpr auto timeout =3D std::named_parameter<int, `timeout_`>;
constexpr auto tries =3D std::named_parameter<int, `tries_`>;

Still trying to find the time to write the proposal for that...


Best,

Roland


or (the most conservative way to rewrite our existing code)

=C2=A0 =C2=A0 static constexpr int default_buffer_size =3D 1= 024;
=C2=A0 =C2=A0 static constexpr int default_timeout =3D 10;
=C2=A0 =C2=A0 static constexpr int default_tries =3D 3;
=C2=A0 =C2=A0 C() : buffer_size_(default_buffer_size), timeout_(default_timeout), tries_(default_tries) {}
=C2=A0 =C2=A0 C(int buffer_size, int timeout, int tries)= : buffer_size_(buffer_size), timeout_(timeout), tries_(tries) {}

=C2=A0 =C2=A0 auto cptr =3D new C(); =C2=A0// usually the de= faults are okay
=C2=A0 =C2=A0 auto c2ptr =3D new C(2048, C::default_timeou= t, C::default_tries); =C2=A0// use a bigger buffer
=C2=A0 =C2=A0 auto c3ptr =3D new C(C::default_buffer_size, 1, C::default_tries); =C2=A0// use the default buffer size but with a shorter timeout
=C2=A0 =C2=A0 auto c4ptr =3D new C(2048, C::default_timeou= t, 5); =C2=A0// testing: bump up the number of tries

I don't really want to make it easier for people to write the "function taking tons of defaulted parameters" antipattern.
Let's teach people how to avoid the use of defaulted parameters, instead.

=E2=80=93Arthur


On Monday, August 17, 2015 at 11:58:41 AM UTC-7, Bo Persson wrote:
On 2015-08-17 20:37, Arthur Tchaikovsky wrote:
> I would really appreciate some feedback from people, or is this proposal
> so uninteresting that there is no point in wasting keyboard on it?

You might wonder how widely used it will be?

If not all parameters are of the same type, you can already get some of
the effect with a few overloads:

=C2=A0 =C2=A0 =C2=A0void f(int a =3D 0, string b =3D "1",
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 complex<double>= c =3D 2.0, type4 d =3D 3);

=C2=A0 =C2=A0 =C2=A0inline void f(int a, complex<double> = c)
=C2=A0 =C2=A0 =C2=A0{ f(a, "1", c, 3); }

=C2=A0 =C2=A0 =C2=A0inline void f(string b, complex<double&g= t; c)
=C2=A0 =C2=A0 =C2=A0{ f(0, b, c, 3); }

assuming that d is put last because you hardly ever want to change the
default.

Just saying that you can already do this without a language change.


=C2=A0 =C2=A0 Bo Persson



>
> On Saturday, 15 August 2015 17:49:50 UTC+1, Arthur Tchaikovsky wrote:
>
> =C2=A0 =C2=A0 Hi,
> =C2=A0 =C2=A0 Does anyone see helpfulness of such solution= :
> =C2=A0 =C2=A0 Having fnc:
>
> =C2=A0 =C2=A0 void f(int a =3D 0, int b =3D 1, int c =3D 2= , int d =3D 3);
>
> =C2=A0 =C2=A0 I think that it would be nice to be able to = say whilst calling it
> =C2=A0 =C2=A0 and requiring only some of those args to be non-default:
>
> =C2=A0 =C2=A0 f(default,2,default,4);
>
> =C2=A0 =C2=A0 Thoughts?
>
>


--

---
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/.

--

---
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>.
--------------010403090403020403070202-- .