www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - const property don't wont return const reference to value

reply "Zhenya" <zheny list.ru> writes:
Hi!
Explain me please what's wrong with this code:

module main;

struct foo
{
	int m_bar;
	 property const ref int bar() const
	{
		return m_bar;
	}
}

void main()
{
}

Error: cast(int)this.m_bar is not an lvalue
Dec 23 2012
parent reply Rainer Schuetze <r.sagitario gmx.de> writes:
On 23.12.2012 14:20, Zhenya wrote:

       property const ref int bar() const


The first const does not bind to the return type, but to the whole 
declaration, so it does the same as the const at the end. You should use

      property ref const(int) bar() const
Dec 23 2012
next sibling parent "Zhenya" <zheny list.ru> writes:
On Sunday, 23 December 2012 at 13:37:33 UTC, Rainer Schuetze 
wrote:

 On 23.12.2012 14:20, Zhenya wrote:

      property const ref int bar() const


 The first const does not bind to the return type, but to the 
 whole declaration, so it does the same as the const at the end. 
 You should use

      property ref const(int) bar() const


Thank you very much!
Dec 23 2012
prev sibling parent reply "Namespace" <rswhite4 googlemail.com> writes:
On Sunday, 23 December 2012 at 13:37:33 UTC, Rainer Schuetze 
wrote:

 On 23.12.2012 14:20, Zhenya wrote:

      property const ref int bar() const


 The first const does not bind to the return type, but to the 
 whole declaration, so it does the same as the const at the end. 
 You should use

      property ref const(int) bar() const


Shouldn't we fix this? Most of the Newcomer are confused with the 
fact, that there are two ways to define a function/method as 
const and that they have to write const(ReturnType) and not const 
ReturnType.
Dec 23 2012
next sibling parent reply "bearophile" <bearophileHUGS lycos.com> writes:
Namespace:


 Shouldn't we fix this? Most of the Newcomer are confused with 
 the fact,


Take a look in bugzilla. I opened an enhancement request to fix 
this situation, but Walter has closed it down. The rationale is 
to keep uniform the way D manages tags like const, immutable, etc.

I don't agree with him in this case: I agree that uniformity is 
good, but breaking symmetry is acceptable when it just produces 
an error and when it helps avoid a common mistake.

If you have strong desires about fixing this, then I suggest you 
to open a new thread about it in the main D newsgroup. I have 
already fought this battle and I lost. (And I think this time 
Jonathan was on my side).

Bye,
bearophile
Dec 23 2012
parent reply "Namespace" <rswhite4 googlemail.com> writes:
On Sunday, 23 December 2012 at 14:55:30 UTC, bearophile wrote:

 Namespace:


 Shouldn't we fix this? Most of the Newcomer are confused with 
 the fact,


 Take a look in bugzilla. I opened an enhancement request to fix 
 this situation, but Walter has closed it down. The rationale is 
 to keep uniform the way D manages tags like const, immutable, 
 etc.

 I don't agree with him in this case: I agree that uniformity is 
 good, but breaking symmetry is acceptable when it just produces 
 an error and when it helps avoid a common mistake.

 If you have strong desires about fixing this, then I suggest 
 you to open a new thread about it in the main D newsgroup. I 
 have already fought this battle and I lost. (And I think this 
 time Jonathan was on my side).

 Bye,
 bearophile


I have to write "const (ReturnType)"  and "ref ReturnType" ... 
very symetric. But I think that Walter thinks that this is the 
best idea, and of course better than that what C++ and other 
languages do​​. So what should I waste my time with "war" with my 
limited english knowledge ? ;)
But thank you, that you tried it.

Maybe this is something for Romulus... :)
Dec 23 2012
next sibling parent "bearophile" <bearophileHUGS lycos.com> writes:
Namespace:


 So what should I waste my time with "war" with my limited 
 english knowledge ? ;)
 But thank you, that you tried it.


Your English improves if you want it, and you exercise yourself.

And such "wars" are not always useless, because while Walter is 
very experienced and usually strong willed, he isn't always 
right. In several situations we/I have eventually changed his 
mind for the better.

Bye,
bearophile
Dec 23 2012
prev sibling parent reply "monarch_dodra" <monarchdodra gmail.com> writes:
On Sunday, 23 December 2012 at 15:05:16 UTC, Namespace wrote:

 On Sunday, 23 December 2012 at 14:55:30 UTC, bearophile wrote:

 Namespace:


 Shouldn't we fix this? Most of the Newcomer are confused with 
 the fact,


 Take a look in bugzilla. I opened an enhancement request to 
 fix this situation, but Walter has closed it down. The 
 rationale is to keep uniform the way D manages tags like 
 const, immutable, etc.

 I don't agree with him in this case: I agree that uniformity 
 is good, but breaking symmetry is acceptable when it just 
 produces an error and when it helps avoid a common mistake.

 If you have strong desires about fixing this, then I suggest 
 you to open a new thread about it in the main D newsgroup. I 
 have already fought this battle and I lost. (And I think this 
 time Jonathan was on my side).

 Bye,
 bearophile


 I have to write "const (ReturnType)"  and "ref ReturnType" ... 
 very symetric. But I think that Walter thinks that this is the 
 best idea, and of course better than that what C++ and other 
 languages do​​. So what should I waste my time with "war" with 
 my limited english knowledge ? ;)
 But thank you, that you tried it.

 Maybe this is something for Romulus... :)


"ref" is a function qualifier, not a type qualifier, so you could 
even write it as:
"const(ReturnType) foo() const ref;"

So technically, it is symetric.

I for one don't think this is a huge problem. There is *some* 
confusion for those comming from C++, but they (we) have to learn 
D is not C++. Besides, I like the liberty of putting all 
qualifiers on the line before the types, including const.

If anything "const" *itself* is VERY confusing for new commers. 
WAY more than the syntax used to qualify a function with it.
Dec 24 2012
parent reply Jonathan M Davis <jmdavisProg gmx.com> writes:
On Monday, December 24, 2012 14:32:55 monarch_dodra wrote:

 "ref" is a function qualifier, not a type qualifier, so you could
 even write it as:
 "const(ReturnType) foo() const ref;"
 
 So technically, it is symetric.


I'm pretty sure that ref is nonsensical in that example. What would ref on a 
function even mean? It could be used on a function pointer, but then you'd 
have to have the function keyword in there. It makes no sense on a function 
prototype like you have there. If that compiles, it's because dmd allows you 
to put all kinds of modifiers on symbols where the modifier has no effect and 
arguably shouldn't be legal.


 I for one don't think this is a huge problem. There is *some*
 confusion for those comming from C++, but they (we) have to learn
 D is not C++. Besides, I like the liberty of putting all
 qualifiers on the line before the types, including const.
 
 If anything "const" *itself* is VERY confusing for new commers.
 WAY more than the syntax used to qualify a function with it.


You're likely to be told to move the const to the right if one of us notices 
it on the left in a pull request. I believe that quite a few of us consider it 
to be bad practice to put it on the left.

- Jonathan M Davis
Dec 24 2012
parent reply "Monarch Dodra" <monarchdodra gmail.com> writes:
On Monday, 24 December 2012 at 15:52:01 UTC, Jonathan M Davis 
wrote:

 On Monday, December 24, 2012 14:32:55 monarch_dodra wrote:

 "ref" is a function qualifier, not a type qualifier, so you 
 could
 even write it as:
 "const(ReturnType) foo() const ref;"
 
 So technically, it is symetric.


 I'm pretty sure that ref is nonsensical in that example. What 
 would ref on a
 function even mean? It could be used on a function pointer, but 
 then you'd
 have to have the function keyword in there. It makes no sense 
 on a function
 prototype like you have there. If that compiles, it's because 
 dmd allows you
 to put all kinds of modifiers on symbols where the modifier has 
 no effect and
 arguably shouldn't be legal.


Wait, never mind that. I thought that was legal.

I was mentioning the fact that if you want to know if a function 
returns ref, you have check that function's attributes:

http://dlang.org/phobos/std_traits.html#FunctionAttribute


 I for one don't think this is a huge problem. There is *some*
 confusion for those comming from C++, but they (we) have to 
 learn
 D is not C++. Besides, I like the liberty of putting all
 qualifiers on the line before the types, including const.
 
 If anything "const" *itself* is VERY confusing for new commers.
 WAY more than the syntax used to qualify a function with it.


 You're likely to be told to move the const to the right if one 
 of us notices
 it on the left in a pull request. I believe that quite a few of 
 us consider it
 to be bad practice to put it on the left.

 - Jonathan M Davis


That's the way the ddoc is generated anyways: Everything on the 
left.
Dec 24 2012
parent Jonathan M Davis <jmdavisProg gmx.com> writes:
On Monday, December 24, 2012 17:25:05 Monarch Dodra wrote:

 That's the way the ddoc is generated anyways: Everything on the
 left.


Which should probably be fixed IMHO. Regardless, in Phobos or druntime, if 
const is found on the left, it gets moved to the right.

- Jonathan M Davis
Dec 24 2012
prev sibling parent reply Jonathan M Davis <jmdavisProg gmx.com> writes:
On Sunday, December 23, 2012 15:41:09 Namespace wrote:

 On Sunday, 23 December 2012 at 13:37:33 UTC, Rainer Schuetze
 
 wrote:

 On 23.12.2012 14:20, Zhenya wrote:

      property const ref int bar() const


 
 The first const does not bind to the return type, but to the
 whole declaration, so it does the same as the const at the end.
 You should use
 
      property ref const(int) bar() const


 
 Shouldn't we fix this? Most of the Newcomer are confused with the
 fact, that there are two ways to define a function/method as
 const and that they have to write const(ReturnType) and not const
 ReturnType.


No one has been able to convince Walter. He thinks that the consistency of 
allowing function attributes on both sides trumps fixing the problems caused by 
having const or immutable on the left.

- Jonathan M Davis
Dec 23 2012
parent reply "Namespace" <rswhite4 googlemail.com> writes:
 No one has been able to convince Walter. He thinks that the 
 consistency of
 allowing function attributes on both sides trumps fixing the 
 problems caused by
 having const or immutable on the left.

 - Jonathan M Davis


It is more confusing than anything else but as I said I would 
waste my time with convince Walter.
What is your opinion?
Dec 23 2012
parent Jonathan M Davis <jmdavisProg gmx.com> writes:
On Sunday, December 23, 2012 21:34:25 Namespace wrote:

 No one has been able to convince Walter. He thinks that the
 consistency of
 allowing function attributes on both sides trumps fixing the
 problems caused by
 having const or immutable on the left.
 
 - Jonathan M Davis


 
 It is more confusing than anything else but as I said I would
 waste my time with convince Walter.
 What is your opinion?


On which? Whether it should be changed or whether it's worth your time trying 
to convince Walter? I'd definitely argue that putting const and immutable on 
the left should be made illegal. The consistency is not worth the problems 
that it causes. But I doubt that there's much hope of convincing Walter at 
this point. The fact that newbies screw it up all the time clearly isn't 
enough to convince him, or he would have relented in the past. And that's 
really the only argument AFAIK, since if you know what you're doing, it's 
trivial to avoid the problem.

- Jonathan M Davis
Dec 23 2012