How to extract mixed format using istringstream

Question

Why does my program not output:

10
1.546
,Apple 1

instead of

10
1
<empty space>

here's my program:

#include <iostream>
#include <string>
#include <sstream>

using namespace std;

int main () {
    string str = "10,1.546,Apple 1";
    istringstream stream (str);
    int a;
    double b;
    string c, dummy;
    stream >> a >> dummy >> b >> dummy >> c;
    cout << a << endl;
    cout << b << endl;
    cout << c << endl;
    return 0;
}

Basically I am trying to parse the comma-separated strings, any smoother way to do this would help me immense.

Answer 1

In IOStreams, strings (meaning both C-strings and C++ strings) have virtually no formatting requirements. Any and all characters are extracted into a string only until a whitespace character is found, or until the end of the stream is caught. In your example, you're using a string intended to eat up the commas between the important data, but the output you are experiencing is the result of the behavior I just explained: The dummy string doesn't just eat the comma, but also the rest of the character sequence until the next whitespace character.

To avoid this you can use a char for the dummy variable, which only has space for one character. And if you're looking to put Apple 1 into a string you will need an unformatted extraction because the formatted extractor operator>>() only reads until whitespace. The appropriate function to use here is std::getline():

string c;
char dummy;

if ((stream >> a >> dummy >> b >> dummy) &&
     std::getline(stream >> std::ws, s))
//   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
{

}

Clearing the newline after the formatted extraction is also necessary which is why I used std::ws to clear leading whitespace. I'm also using an if statement to contain the extraction in order to tell if it succeeded or not.

---

Any smoother way to do this would help me immensely.

You can set the classification of the comma character to a whitespace character using the std::ctype<char> facet of the locale imbued in the stream. This will make the use of a dummy variable unnecessary. Here's an example:

namespace detail
{
    enum options { add, remove };

    class ctype : public std::ctype<char>
    {
    private:
        static mask* get_table(const std::string& ws, options opt)
        {
            static std::vector<mask> table(classic_table(),
                                           classic_table() + table_size);
            for (char c : ws)
            {
                if (opt == add)
                    table[c] |= space;
                else if (opt == remove)
                    table[c] &= ~space;
            }
            return &table[0];
        }
    public:
        ctype(const std::string& ws, options opt)
            : std::ctype<char>(get_table(ws, opt)) { }
    };
}

class adjustws_impl
{
public:
    adjustws_impl(const std::string& ws, detail::options opt) :
        m_ws(ws),
        m_opt(opt)
    { }

    friend std::istream& operator>>(std::istream& is,
                                    const adjustws_impl& manip)
    {
        const detail::ctype* facet(new detail::ctype(manip.m_ws, manip.m_opt));

        if (!std::has_facet<detail::ctype>(is.getloc())
        {
            is.imbue(std::locale(is.getloc(), facet));
        } else
            delete facet;

        return is;
    }
private:
    std::string m_ws;
    detail::options m_opt;
};

adjustws_impl setws(const std::string& ws)
{
    return adjustws_impl(ws, detail::add);
}

adjustws_impl unsetws(const std::string& ws)
{
    return adjustws_impl(ws, detail::remove);
}

int main()
{
    std::istringstream iss("10,1.546,Apple 1");
    int a; double b; std::string c;

    iss >> setws(","); // set comma to a whitespace character

    if ((iss >> a >> b) && std::getline(iss >> std::ws, c))
    {
        // ...
    }

    iss >> unsetws(","); // remove the whitespace classification
}