Re: [boost] PR: Remove safe_bool idiom from boost.tribool

Subject: Re: [boost] PR: Remove safe_bool idiom from boost.tribool
From: Robert Ramey (ramey_at_[hidden])
Date: 2018-05-22 15:30:42

>> Finally.  I believe that these are compile time errors in C++17.
> 
> No, they are not.
> 
> struct tribool
> {
>  Â Â Â  int value;
> 
>  Â Â Â  EXPLICIT operator bool() const { return value != 0; }
> };
> 
> int main()
> {
>  Â Â Â  tribool b1, b2;
>  Â Â Â  b1 < b2;
>  Â Â Â  b1 + b2;
>  Â Â Â  b1 + 5;
>  Â Â Â  b1 * 2;
> }
> 
> $ g++ -std=c++17 -DEXPLICIT= -o tribool_conv tribool_conv.cpp
> 
> $ g++ -std=c++17 -DEXPLICIT=explicit -o tribool_conv tribool_conv.cpp
> tribool_conv.cpp: In function ‘int main()’:
> tribool_conv.cpp:11:8: error: no match for ‘operator<’ (operand types 
> are ‘tribool’ and ‘tribool’)
>  Â Â Â Â  b1 < b2;
>  Â Â Â Â  ~~~^~~~
> tribool_conv.cpp:11:8: note: candidate: operator<(int, int) <built-in>
> tribool_conv.cpp:11:8: note:Â Â  no known conversion for argument 2 from 
> ‘tribool’ to ‘int’
> tribool_conv.cpp:12:8: error: no match for ‘operator+’ (operand types 
> are ‘tribool’ and ‘tribool’)
>  Â Â Â Â  b1 + b2;
>  Â Â Â Â  ~~~^~~~
> tribool_conv.cpp:12:8: note: candidate: operator+(int, int) <built-in>
> tribool_conv.cpp:12:8: note:Â Â  no known conversion for argument 2 from 
> ‘tribool’ to ‘int’
> tribool_conv.cpp:13:8: error: no match for ‘operator+’ (operand types 
> are ‘tribool’ and ‘int’)
>  Â Â Â Â  b1 + 5;
>  Â Â Â Â  ~~~^~~
> tribool_conv.cpp:13:8: note: candidate: operator+(int, int) <built-in>
> tribool_conv.cpp:13:8: note:Â Â  no known conversion for argument 1 from 
> ‘tribool’ to ‘int’
> tribool_conv.cpp:14:8: error: no match for ‘operator*’ (operand types 
> are ‘tribool’ and ‘int’)
>  Â Â Â Â  b1 * 2;
>  Â Â Â Â  ~~~^~~
> tribool_conv.cpp:14:8: note: candidate: operator*(int, int) <built-in>
> tribool_conv.cpp:14:8: note:Â Â  no known conversion for argument 1 from 
> ‘tribool’ to ‘int’

Damn, you got me here. Good illustration. So the legal operations of a 
bool depends on whether it's the result of an explicit or implicit 
conversion. Such a think would never have occurred to me.  I suppose 
it's obvious what I think about that idea so I won't comment.

Another confession: I didn't realize that optional used and explicit 
conversion.  Maybe that changes something.  I've never thought that 
conversion to bool from optional was ever a good idea - be it implicit 
or explicit.  But I guess that's just me.

Robert Ramey