Re: [boost] Bug in boost::rational, no Boolean conversion

From: Jonathan Turkanis (technews_at_[hidden])
Date: 2005-07-17 16:34:39

Daryle Walker wrote:
> On 7/17/05 1:57 AM, "Jonathan Turkanis" <technews_at_[hidden]>
> wrote:
>
>> Daryle Walker wrote:
>>> I'm writing a class template where the template parameter is a type
>>> that should represent real numbers.  It started choking when using
>>> boost::rational when I needed Boolean conversion, [...]
>>
>>> Before anyone says that's proper, realize that boost::rational DOES
>>> have "operator not()" defined!  This is a surprise to the typical
>>> C++ programmer.  Either both should be defined, or neither.  I
>>> think that boost::rational should add a Boolean conversion.  (It's
>>> a quick fix, and it is critical, so it should be done for this
>>> release.)
>>
>> Off the top of my head, I can't think why it shouldn't be added.
>> However, there might be a good reason I can't think of, and adding
>> the operator could conceivably break existing code, so I don't think
>> it's appropriate to stick it in a few days before release.
>
> Boolean conversions are currently banned, so they couldn't be in
> existing code.  The worry is expressions where a conversion would be
> higher priority that whatever operation is currently used.  The only
> operations I can think of in that category are other conversions, but
> a new Boolean conversion can't cause problems because C++ allows at
> most one user-conversion (i.e. the new operation is of lower
> priority).  If we use a member-pointer and keep the type & values
> used private, then no one can name the type to grab it directly.  The
> class template already defines equality operators, so we don't have
> to worry about the member-pointer conversion adding inappropriate
> operators, especially if we make each version of boost::rational
> maintain separate Boolean types.

As I said, I can't think why it shouldn't be added.

>> I'll add it to the list of feature requests for 1.34.
>
> I put it in our SourceForge bug tracker as Request ID #1239906, so we
> won't forget.

Okay.

>> In the mean time, maybe you can use != 0 or !!
>
> I have "x != T()" right now, but I want to keep the code clean.
> Especially since I specified Boolean conversion as a requirement.  (I
> didn't know about boost::rational's deficiency at first.)

Jonathan