Re: [boost] boost::rational test program

From: Paul Moore (gustav_at_[hidden])
Date: 2001-01-21 17:33:27

On 15 Jan 2001, at 21:43, Stephen Silver wrote:

> I've written a program that runs a hundred or so simple tests on
> boost::rational.  In addition to the operator< bugs that I posted
> yesterday, this test program found problems with lcm() (sometimes
> returns negative numbers, and lcm(0,0) causes division by zero),
> and with operator>> (input errors not always handled correctly).

lcm bugs fixed. However, I think you are mistaken over the 
operator>> errors - remember that >> consumes as much input as 
it can, only setting the error state if it cannot construct a valid value.

Your example

    std::istringstream iss("57A");
    iss >> r0;

consumes "57". This is a valid representation of a rational, namely 
57/1, so r0 is set to this, and the stream is in a good state.

Similarly,

    std::istringstream iss("20-20");
    iss >> r0;

sets r0 to 20, and leaves "-20" to be read, without signalling an 
error.

I don't see how I could give behaviour consistent with other types, 
unless I disallowed constructing a rational from a stream containing 
a plain integer (which I don't want to do).

Paul.