BNF Question

[Jascha Wetzel]

This one isn't actually a problem. But you will find that there are more 
tricky ambiguities. For example:

void main()
{
   foo.bar.baz;
}

"foo.bar.baz" could be a type,
"foo.bar" could be a type and ".baz" a property,
"foo" could be an object, "bar" a member, and "baz" a member of that,
resulting in 3 different syntax trees:
BasicType
   IdentifierList
or
PrimaryExpression
   BasicType . Identifier
or
PostfixExpression
   PostfixExpression . PrimaryExpression

The differences are only semantical.

Kirk McDonald wrote:
> BCS wrote:
>> Reply to Xinok,
>>
>>> An equal expression is when the data in two objects is compared.
>>> int[] arr1 = [10, 20, 30], arr2 = [10, 20, 30];
>>> if(arr1 == arr2){ } // True
>>> An identity expresion is when the pointer values (and .length in
>>> arrays)
>>> in two objects is compared.
>>> int[] arr1 = [10, 20, 30], arr2 = arr1;
>>> if(arr1 is arr2){ } // True
>>> For primitive types, there is no difference between using == and is.
>>> int a = 15, b = 15;
>>> if(a == b){ } // True
>>> if(a is b){ } // True
>>> BCS wrote:
>>>
>>
>> However what you describe is at the semantic level and the grammar is 
>> totally at the syntax level. What I was looking at is that this:
>>
>> a is b
>>
>> can be parsed as an Identity Expression and an Equal Expression.
>>
>> If the distinction is a semantic issue then this makes D context 
>> sensitive. If it is not a semantic issue, then it is redundant or 
>> conflicting.
>>
> 
> The grammar is merely confusing. It might be better written as:
> 
> EqualExpression:
>     ShiftExpression
>     ShiftExpression == ShiftExpression
>     ShiftExpression != ShiftExpression
>     IdentityExpression
>