This is a bug?

import std.stdio;

void main() {
    int a = 0;

    writeln( (a < 10) ? a = 1 : a = 2 );    // prints 2

    writeln( (a < 10) ? a = 1 : (a = 2) );  // prints 1
}

Even C++ output:
1
1

On Friday, 13 February 2015 at 09:38:04 UTC, Dennis Ritchie wrote:
> This is a bug?
>
> import std.stdio;
>
> void main() {
>     int a = 0;
>
>     writeln( (a < 10) ? a = 1 : a = 2 );    // prints 2
>
>     writeln( (a < 10) ? a = 1 : (a = 2) );  // prints 1
> }
>
> Even C++ output:
> 1
> 1

About 2 years ago, I had a problem with similar structure.

My guess is that the first one is accepted as this.

((a < 10) ? a = 1 : a)   =   ( 2 )

Thereby it gives this result. Vague definitions are always error prone.

On 2/13/15 7:38 AM, tcak wrote:
> On Friday, 13 February 2015 at 09:38:04 UTC, Dennis Ritchie wrote:
>> This is a bug?
>>
>> import std.stdio;
>>
>> void main() {
>>     int a = 0;
>>
>>     writeln( (a < 10) ? a = 1 : a = 2 );    // prints 2
>>
>>     writeln( (a < 10) ? a = 1 : (a = 2) );  // prints 1
>> }
>>
>> Even C++ output:
>> 1
>> 1
>
> About 2 years ago, I had a problem with similar structure.
>
> My guess is that the first one is accepted as this.
>
> ((a < 10) ? a = 1 : a)   =   ( 2 )
>
> Thereby it gives this result. Vague definitions are always error prone.

Yes, the operator precedence (curiously not defined in the spec) is here:

http://wiki.dlang.org/Operator_precedence

Conditional operator is above assignment operators.

C++ operator precedence is here:

http://en.wikipedia.org/wiki/Operators_in_C_and_C%2B%2B#Operator_precedence

Note they are the same, but I think C++ ?: = operator in this case does not result in lvalue. So it must be that the = cannot bind to the result of the operator, therefore it binds to the a. Interesting...

-Steve

On Friday, 13 February 2015 at 13:25:55 UTC, Steven Schveighoffer wrote:
> Yes, the operator precedence (curiously not defined in the spec) is here:
>
> http://wiki.dlang.org/Operator_precedence
>
> Conditional operator is above assignment operators.

It's specified through the grammar [1]:

----
AssignExpression:
    ConditionalExpression
    ConditionalExpression = AssignExpression
    [...]

ConditionalExpression:
    OrOrExpression
    OrOrExpression ? Expression : ConditionalExpression
----

The third operand of ConditionalExpression cannot be an AssignExpression without parentheses.

So `foo ? bar : baz = qux` can only be interpreted as `(foo ? bar : baz) = qux`.

[1] http://dlang.org/expression.html

On Friday, 13 February 2015 at 09:38:04 UTC, Dennis Ritchie wrote:
> This is a bug?
>
> import std.stdio;
>
> void main() {
>     int a = 0;
>
>     writeln( (a < 10) ? a = 1 : a = 2 );    // prints 2
>
>     writeln( (a < 10) ? a = 1 : (a = 2) );  // prints 1
> }
>
> Even C++ output:
> 1
> 1

Maybe a similar issue as this: http://forum.dlang.org/thread/wlpmjgbkdzcyuuyafzwk@forum.dlang.org

On Friday, 13 February 2015 at 13:25:55 UTC, Steven Schveighoffer wrote:
> On 2/13/15 7:38 AM, tcak wrote:
>> On Friday, 13 February 2015 at 09:38:04 UTC, Dennis Ritchie wrote:
>>> This is a bug?
>>>
>>> import std.stdio;
>>>
>>> void main() {
>>>    int a = 0;
>>>
>>>    writeln( (a < 10) ? a = 1 : a = 2 );    // prints 2
>>>
>>>    writeln( (a < 10) ? a = 1 : (a = 2) );  // prints 1
>>> }
>>>
>>> Even C++ output:
>>> 1
>>> 1
>>
>> About 2 years ago, I had a problem with similar structure.
>>
>> My guess is that the first one is accepted as this.
>>
>> ((a < 10) ? a = 1 : a)   =   ( 2 )
>>
>> Thereby it gives this result. Vague definitions are always error prone.
>
> Yes, the operator precedence (curiously not defined in the spec) is here:
>
> http://wiki.dlang.org/Operator_precedence
>
> Conditional operator is above assignment operators.
>
> C++ operator precedence is here:
>
> http://en.wikipedia.org/wiki/Operators_in_C_and_C%2B%2B#Operator_precedence
>
> Note they are the same, but I think C++ ?: = operator in this case does not result in lvalue. So it must be that the = cannot bind to the result of the operator, therefore it binds to the a. Interesting...
>
> -Steve

Thanks.

On 2/13/15 9:01 AM, anonymous wrote:
> On Friday, 13 February 2015 at 13:25:55 UTC, Steven Schveighoffer wrote:
>> Yes, the operator precedence (curiously not defined in the spec) is here:
>>
>> http://wiki.dlang.org/Operator_precedence
>>
>> Conditional operator is above assignment operators.
>
> It's specified through the grammar [1]:
>
> ----
> AssignExpression:
>      ConditionalExpression
>      ConditionalExpression = AssignExpression
>      [...]
>
> ConditionalExpression:
>      OrOrExpression
>      OrOrExpression ? Expression : ConditionalExpression
> ----
>
> The third operand of ConditionalExpression cannot be an AssignExpression
> without parentheses.
>
> So `foo ? bar : baz = qux` can only be interpreted as `(foo ? bar : baz)
> = qux`.
>
> [1] http://dlang.org/expression.html

Right, but C++ seems to treat it differently, but has the same order of precedence. It seems that C++ can do lvalues as a result of assignment, but for some reason it doesn't do the same as D.

Indeed, if you do:

(a < 10) ? 5 : a = 1;

In D, this results in an error, in C++, this just results in the rvalue 5.

OK, so I just looked for more information, and found that in some references, the ?: is given the same precedence as the assignment operator (and ordered right to left), which would explain the behavior. So likely the wikipedia reference is just wrong.

e.g.: http://en.cppreference.com/w/cpp/language/operator_precedence

Walter (or other language gurus), any reason that D does it this way? The C++ way seems more natural to me.

-Steve

It is your profile?
http://www.cyberforum.ru/members/491746.html