Suggestion for a function for checking fraction number

[kouisawang]

I have written a function to check if the number is fraction in either X/X or X X/X format where X is double.

Here is the code.

 
bool IsFraction(const CComBSTR bucketSize, double &result)
{    
    int nToken=1;
    HRESULT hr = S_OK;
    CComBSTR tempBucketSize(bucketSize);
    //Count the token. must be 2 or 3
    wcstok(tempBucketSize.m_str, _T(" /"));
    while(wcstok(NULL, _T(" /")))
    {
        ++nToken;
    }
 
    //If number token == 2, the number must be in the form X/X
    if(nToken == 2)
    {
        CComBSTR szToken = wcstok(bucketSize.m_str, _T("/"));
 
        CComVariant vtNumerator = szToken;
        hr = vtNumerator.ChangeType( VT_R8 );
        //if szToken is not number return false
        if(FAILED(hr) || (vtNumerator.vt != VT_R8) || (vtNumerator.dblVal < 0.0))
            return false;
 
        //if there's no string left or the second string is not number or it's zero, return false
        szToken = wcstok(NULL, _T("/"));
        CComVariant vtDenominator = szToken;
        hr = vtDenominator.ChangeType(VT_R8);
 
        if(FAILED(hr) || (vtDenominator.vt != VT_R8) || (vtDenominator.dblVal <= 0.0))
            return false;
 
        //otherwise calculate number and return true
        result = vtNumerator.dblVal / vtDenominator.dblVal;
        return true;
    }
    //If number token == 3, the number must be in the form X X/X
    else if(nToken == 3)
    {
        CComBSTR szToken = wcstok(bucketSize.m_str, _T(" "));
 
        CComVariant vtNumber = szToken;
        hr = vtNumber.ChangeType(VT_R8);
        //if szToken is not number return false
        if(FAILED(hr) || (vtNumber.vt != VT_R8) || (vtNumber.dblVal < 0.0))
            return false;
 
        //if there's no string left or the second string is not number, return false
        szToken = wcstok(NULL, _T("/"));
        CComVariant vtNumerator = szToken;
        hr = vtNumerator.ChangeType(VT_R8);
 
        if(FAILED(hr) || (vtNumerator.vt != VT_R8) || (vtNumerator.dblVal < 0.0))
            return false;
 
        //if there's no string left or the third string is not number or it's zero, return false
        szToken = wcstok(NULL, _T(" /"));
        CComVariant vtDenominator = szToken;
        hr = vtDenominator.ChangeType(VT_R8);
 
        if(FAILED(hr) || (vtDenominator.vt != VT_R8) || (vtDenominator.dblVal <= 0.0))
            return false;
 
        //otherwise calculate number and return true
        result = vtNumber.dblVal + vtNumerator.dblVal / vtDenominator.dblVal;
        return true;
 
    }
    else
        return false;
}
 

It's very long but I can't think of another algorithm to deal with it. So if you have an idea to optimize it, please give me your suggestions or comments.
Thank

[newb16]

Is 1.25 3.67/0.87 a fraction too, like 1 1/4 is ? If yes then ok, without decreasing number of checks it probably can't be reduced. What is VT_R8, btw?

[weaknessforcats]

1 1/4 is not a fraction. It's a mixed number. The actual number is 5/4 which is a rational number.

1.25 3.67/0.87 is incorrect syntax. The 3.67/0.87 is a real number.

Note that 5/4 maps to 5.0/4.0 making the rational numbers a subset of the real numbers.

If you use the form X/X where X is a floating point number, you will have a correct representation.

[donbock]

kouisawang ...

What precise forms do you need to recognize as fractions?

100          whole number (implied denominator of 1)
1/4          ratio of integers
0.45         decimal fraction (denominator implied by number of digits)
2.67         mixed number (decimal fraction)
2 1/2        mixed number (ratio, whitespace separator)
2-1/2        mixed number (ratio, dash separator)
1.5/7.8      ratio of real numbers
1.2 1.5/7.8  funky mixed number
4,5          European decimal point

How many ways do you need to support negative numbers?
-100, -1/4, 1/-4, -2 1/2, -(2 1/2), etc

[kouisawang]

My intention is to have X/X and X X/X as a valid input format. X is double. however, it only supports positive number. Maybe using "fraction" is not right coz I do count numerator number with decimal point as valid input but ... you get my intention.
btw, vt_r8 is a double variant type

[weaknessforcats]

You might consider a VT_I4 or a VT_UI4 rather than a
VT_R8 for your VARIANT.

You can always create a X X/X from a X/X and you can convert an input X X/X to an X/X. So the common format X/X seems you best bet.

Further, numerators and denominators are integers and not reals. Hcne, I recommend the VT_I4 or for positives only, a VT_UI4.