• Temtaime (15/15) May 17 2013 struct A {
• Namespace (3/3) May 17 2013 I don't see where k comes from?
• Maxim Fomin (6/9) May 17 2013 I think the point is that it comes from nowhere. Compiler
• Temtaime (3/3) May 17 2013 Yes, i want to see 'undeclared variable' error, but the compiler
• =?UTF-8?B?QWxpIMOHZWhyZWxp?= (4/7) May 17 2013 "Iden­ti­fiers start­ing with __ (two un­der­scores) are re­served...
• Jonathan M Davis (8/12) May 17 2013 Identifiers begining with two underscores are reserved by the compiler, ...
• bearophile (5/6) May 17 2013 Ideally you are right. But I think sometimes you want to use
• Namespace (2/11) May 17 2013 Ok, not always, but mostly it's a bad idea.
• 1100110 (3/14) May 17 2013 It is not the D way to forbid you from shooting yourself in the foot.
• Jonathan M Davis (11/17) May 17 2013 D will stop you from shooting yourself in the foot if it can do so in a ...
• bearophile (7/18) May 17 2013 I think to answer this issue we need someone that knows more
• Jonathan M Davis (5/9) May 17 2013 Oh, I'm sure that it would be possible. It's just a question of how
• 1100110 (4/15) May 17 2013 I love how we are all basically agreeing with one another, even though
• Timon Gehr (2/17) May 18 2013 It is trivial to catch them in the parser.
• bearophile (4/5) May 18 2013 So the question is: is it right to catch them all? :-)
• =?UTF-8?B?QWxpIMOHZWhyZWxp?= (8/10) May 18 2013 I don't think so. Double underscores are reserved for the
• Maxim Fomin (7/35) May 17 2013 The only scenario when __identifiers are needed which came to my
• Jesse Phillips (6/7) May 17 2013 I'm pretty sure the answer is:
• bearophile (8/12) May 17 2013 It's a bad error message, but it's correct, because k doesn't

"Temtaime" <temtaime gmail.com> writes:

struct A {
	int opBinary(string op)(A) { k++; return 0; }
}

struct B {
	A __a;
	alias __a this;
}

void main() {
     B b;
     auto c = b * b;
}


This code doesn't compiles with an error:
Error: 'b' is not of arithmetic type, it is a B

If i remove k++, then it's OK. It seems that compiler omitting 
the function if there an error in it.

"Namespace" <rswhite4 googlemail.com> writes:

I don't see where k comes from?
And it's always a bad idea to prefix a variable with '__', 
because this is reserved to the compiler.

"Maxim Fomin" <maxim maxim-fomin.ru> writes:

On Friday, 17 May 2013 at 14:12:10 UTC, Namespace wrote:
 I don't see where k comes from?

I think the point is that it comes from nowhere. Compiler 
incorectly omits two errors from output: "Error undefined 
indentifier" and "template instance error instantiation". This 
should go to bugzilla.

 And it's always a bad idea to prefix a variable with '__', 
 because this is reserved to the compiler.

If it alwalys a bad idea, compiler should not accept such code.

"Temtaime" <temtaime gmail.com> writes:

Yes, i want to see 'undeclared variable' error, but the compiler 
omits function declaration instead.

And why i shouldn't use __? I think it's ordinary identifier.

=?UTF-8?B?QWxpIMOHZWhyZWxp?= <acehreli yahoo.com> writes:

On 05/17/2013 07:40 AM, Temtaime wrote:
 Yes, i want to see 'undeclared variable' error, but the compiler omits
 function declaration instead.

 And why i shouldn't use __? I think it's ordinary identifier.

"Iden­ti­fiers start­ing with __ (two un­der­scores) are re­served."

   http://dlang.org/lex.html#Identifier

Ali

"Jonathan M Davis" <jmdavisProg gmx.com> writes:

On Friday, May 17, 2013 16:40:23 Temtaime wrote:
 Yes, i want to see 'undeclared variable' error, but the compiler
 omits function declaration instead.
 
 And why i shouldn't use __? I think it's ordinary identifier.

Identifiers begining with two underscores are reserved by the compiler, which 
is why identifiers such as __FILE__, __LINE__, __traits, and __gshared all 
begin with two underscores. Identifiers that the compiler generates are going 
to start with two underscores, and you could end up with bugs in your program 
if you happened to declare an identifier with a name that matched a compiler 
generated one. I believe that that's the case with most C-based languages.

- Jonathan M Davis

"bearophile" <bearophileHUGS lycos.com> writes:

Maxim Fomin:

 If it alwalys a bad idea, compiler should not accept such code.

Ideally you are right. But I think sometimes you want to use 
those reserved names... So I don't know.

Bye,
bearophile

"Namespace" <rswhite4 googlemail.com> writes:

On Friday, 17 May 2013 at 14:35:35 UTC, Maxim Fomin wrote:
 On Friday, 17 May 2013 at 14:12:10 UTC, Namespace wrote:
 I don't see where k comes from?

 I think the point is that it comes from nowhere. Compiler 
 incorectly omits two errors from output: "Error undefined 
 indentifier" and "template instance error instantiation". This 
 should go to bugzilla.

 And it's always a bad idea to prefix a variable with '__', 
 because this is reserved to the compiler.

 If it alwalys a bad idea, compiler should not accept such code.

Ok, not always, but mostly it's a bad idea.

1100110 <0b1100110 gmail.com> writes:

On 05/17/2013 09:35 AM, Maxim Fomin wrote:
 On Friday, 17 May 2013 at 14:12:10 UTC, Namespace wrote:
 I don't see where k comes from?

=20
 I think the point is that it comes from nowhere. Compiler incorectly
 omits two errors from output: "Error undefined indentifier" and
 "template instance error instantiation". This should go to bugzilla.
=20
 And it's always a bad idea to prefix a variable with '__', because
 this is reserved to the compiler.

=20
 If it alwalys a bad idea, compiler should not accept such code.

It is not the D way to forbid you from shooting yourself in the foot.

goto, catch(Throwable th), there are plenty more examples.

"Jonathan M Davis" <jmdavisProg gmx.com> writes:

On Friday, May 17, 2013 11:59:38 1100110 wrote:
 On 05/17/2013 09:35 AM, Maxim Fomin wrote:
 If it alwalys a bad idea, compiler should not accept such code.

 
 It is not the D way to forbid you from shooting yourself in the foot.
 
 goto, catch(Throwable th), there are plenty more examples.

D will stop you from shooting yourself in the foot if it can do so in a way 
that doesn't actually limit you, and it does do quite a few things to make it 
harder to shoot yourself in the foot, but it certainly doesn't ultimately stop 
you from doing so. However, if it _never_ makes sense to declare a variable 
beginning with two underscores, I don't know why the compiler wouldn't forbid 
it other than the fact that it probably inserts such variables prior to when 
it would do the semantic analysis to check whether variables started with 
underscores, in which case, catching the user's variables that start with two 
underscores while permitting the compiler's variables could get tricky.

- Jonathan M Davis

"bearophile" <bearophileHUGS lycos.com> writes:

Jonathan M Davis:

 However, if it _never_ makes sense to declare a variable
 beginning with two underscores, I don't know why the compiler 
 wouldn't forbid
 it other than the fact that it probably inserts such variables 
 prior to when
 it would do the semantic analysis to check whether variables 
 started with
 underscores, in which case, catching the user's variables that 
 start with two
 underscores while permitting the compiler's variables could get 
 tricky.

I think to answer this issue we need someone that knows more 
about the DMD compiler (as Kenji Hara). I think that if we want 
to forbid those variables in user code, then probably there is a 
way do to it.

Bye,
bearophile

"Jonathan M Davis" <jmdavisProg gmx.com> writes:

On Friday, May 17, 2013 19:19:46 bearophile wrote:
 I think to answer this issue we need someone that knows more
 about the DMD compiler (as Kenji Hara). I think that if we want
 to forbid those variables in user code, then probably there is a
 way do to it.

Oh, I'm sure that it would be possible. It's just a question of how 
complicated that would be and whether it's worth it given that very few 
programmers try and declare variables which start with two underscores.

- Jonathan M Davis

1100110 <0b1100110 gmail.com> writes:

On 05/17/2013 12:27 PM, Jonathan M Davis wrote:
 On Friday, May 17, 2013 19:19:46 bearophile wrote:
 I think to answer this issue we need someone that knows more
 about the DMD compiler (as Kenji Hara). I think that if we want
 to forbid those variables in user code, then probably there is a
 way do to it.

=20
 Oh, I'm sure that it would be possible. It's just a question of how=20
 complicated that would be and whether it's worth it given that very few=

=20
 programmers try and declare variables which start with two underscores.=

=20
 - Jonathan M Davis


I love how we are all basically agreeing with one another, even though
we are saying such different things.

Timon Gehr <timon.gehr gmx.ch> writes:

On 05/17/2013 07:19 PM, bearophile wrote:
 Jonathan M Davis:

 However, if it _never_ makes sense to declare a variable
 beginning with two underscores, I don't know why the compiler wouldn't
 forbid
 it other than the fact that it probably inserts such variables prior
 to when
 it would do the semantic analysis to check whether variables started with
 underscores, in which case, catching the user's variables that start
 with two
 underscores while permitting the compiler's variables could get tricky.

 I think to answer this issue we need someone that knows more about the
 DMD compiler (as Kenji Hara). I think that if we want to forbid those
 variables in user code, then probably there is a way do to it.

 Bye,
 bearophile

It is trivial to catch them in the parser.

"bearophile" <bearophileHUGS lycos.com> writes:

Timon Gehr:

 It is trivial to catch them in the parser.

So the question is: is it right to catch them all? :-)

Bye,
bearophile

=?UTF-8?B?QWxpIMOHZWhyZWxp?= <acehreli yahoo.com> writes:

On 05/18/2013 11:50 AM, bearophile wrote:> Timon Gehr:
 It is trivial to catch them in the parser.

 So the question is: is it right to catch them all? :-)

I don't think so. Double underscores are reserved for the 
"implementation", which includes the standard library. I don't think the 
compiler should know that it is compiling something that is part of Phobos.

It could even be a mixin that has such a variable that is implemented in 
Phobos. There is nothing wrong in reminder each other that a name is 
reserved. :)

Ali

"Maxim Fomin" <maxim maxim-fomin.ru> writes:

On Friday, 17 May 2013 at 17:08:31 UTC, Jonathan M Davis wrote:
 On Friday, May 17, 2013 11:59:38 1100110 wrote:
 On 05/17/2013 09:35 AM, Maxim Fomin wrote:
 If it alwalys a bad idea, compiler should not accept such 
 code.

 
 It is not the D way to forbid you from shooting yourself in 
 the foot.
 
 goto, catch(Throwable th), there are plenty more examples.

 D will stop you from shooting yourself in the foot if it can do 
 so in a way
 that doesn't actually limit you, and it does do quite a few 
 things to make it
 harder to shoot yourself in the foot, but it certainly doesn't 
 ultimately stop
 you from doing so. However, if it _never_ makes sense to 
 declare a variable
 beginning with two underscores, I don't know why the compiler 
 wouldn't forbid
 it other than the fact that it probably inserts such variables 
 prior to when
 it would do the semantic analysis to check whether variables 
 started with
 underscores, in which case, catching the user's variables that 
 start with two
 underscores while permitting the compiler's variables could get 
 tricky.

 - Jonathan M Davis

The only scenario when __identifiers are needed which came to my 
mind is case when working with C or os implementation. Low-level 
implementation supposed to use '__' namespace because of practice 
encouraged by ISO C and other docs to reserve space for user and 
system specific identifiers. Probably double leading underscores 
should be restricted to extern declarations.

"Jesse Phillips" <jessek.phillips+D gmail.m> writes:

On Friday, 17 May 2013 at 14:35:35 UTC, Maxim Fomin wrote:
 If it alwalys a bad idea, compiler should not accept such code.

I'm pretty sure the answer is:

Compilers can have different extensions, it would be bad for a 
compiler to reject code which is valid by another compiler.

That said, I don't know of any rules around ignoring these as 
there is for pragma.

"bearophile" <bearophileHUGS lycos.com> writes:

Temtaime:

 This code doesn't compiles with an error:
 Error: 'b' is not of arithmetic type, it is a B

 If i remove k++, then it's OK. It seems that compiler omitting 
 the function if there an error in it.

It's a bad error message, but it's correct, because k doesn't 
exists. This error message is so bad because opBinary is a 
template.

In Bugzilla there is a request for an improvement related to this:
http://d.puremagic.com/issues/show_bug.cgi?id=9715

Bye,
bearophile