c++ Reading numbers from text files, ignoring comments

Question

So I've seen lots of solutions on this site and tutorials about reading in from a text file in C++, but have yet to figure out a solution to my problem. I'm new at C++ so I think I'm having trouble piecing together some of the documentation to make sense of it all.

What I am trying to do is read a text file numbers while ignoring comments in the file that are denoted by "#". So an example file would look like:

#here is my comment
20 30 40 50
#this is my last comment
60 70 80 90

My code can read numbers fine when there aren't any comments, but I don't understand parsing the stream well enough to ignore the comments. Its kind of a hack solution right now.

/////////////////////// Read the file ///////////////////////
std::string line;
if (input_file.is_open())
{
    //While we can still read the file
    while (std::getline(input_file, line))
    {
        std::istringstream iss(line);
        float num; // The number in the line

        //while the iss is a number 
        while ((iss >> num))
        {
            //look at the number
        }
    }
}

else
{
    std::cout << "Unable to open file";
}
/////////////////////// done reading file /////////////////

Is there a way I can incorporate comment handling with this solution or do I need a different approach? Any advice would be great, thanks.

Answer 1

If your file contains # always in the first column, then just test, if the line starts with # like this:

while (std::getline(input_file, line))
{
    if (line[0] != "#" )
    {
        std::istringstream iss(line);
        float num; // The number in the line

        //while the iss is a number 
        while ((iss >> num))
        {
            //look at the number
        }
    }
}

It is wise though to trim the line of leading and trailing whitespaces, like shown here for example: Remove spaces from std::string in C++

Answer 2

If this is just a one of use, for line oriented input like yours, the simplest solution is just to strip the comment from the line you just read:

line.erase( std::find( line.begin(), line.end(), '#' ), line.end() );

A more generic solution would be to use a filtering streambuf, something like:

class FilterCommentsStreambuf : public std::streambuf
{
    std::istream& myOwner;
    std::streambuf* mySource;
    char myCommentChar;
    char myBuffer;

protected:
    int underflow()
    {
        int const eof = std::traits_type::eof();
        int results = mySource->sbumpc();
        if ( results == myCommentChar ) {
            while ( results != eof && results != '\n') {
                results = mySource->sbumpc(0;
            }
        }
        if ( results != eof ) {
            myBuffer = results;
            setg( &myBuffer, &myBuffer, &myBuffer + 1 );
        }
        return results;
    }

public:
    FilterCommentsStreambuf( std::istream& source,
                             char comment = '#' )
        : myOwner( source )
        , mySource( source.rdbuf() )
        , myCommentChar( comment )
    {
        myOwner.rdbuf( this );
    }
    ~FilterCommentsStreambuf()
    {
        myOwner.rdbuf( mySource );
    }
};

In this case, you could even forgo getline:

FilterCommentsStreambuf filter( input_file );
double num;
while ( input_file >> num || !input_file.eof() ) {
    if ( ! input_file ) {
        //  Formatting error, output error message, clear the
        //  error, and resynchronize the input---probably by
        //  ignore'ing until end of line.
    } else {
        //  Do something with the number...
    }
}

(In such cases, I've found it useful to also track the line number in the FilterCommentsStreambuf. That way you have it for error messages.)

Answer 3

An alternative to the "read aline and parse it as a string", can be use the stream itself as the incoming buffer:

while(input_file)
{
    int n = 0;

    char c; 
    input_file >> c; // will skip spaces ad read the first non-blank

    if(c == '#')
    {
        while(c!='\n' && input_file) input_file.get(c);
        continue; //may be not soooo beautiful, but does not introduce useless dynamic memory
    }

    //c is part of something else but comment, so give it back to parse it as number
    input_file.unget(); //< this is what all the fuss is about!
    if(input_file >> n)
    { 
        // look at the nunber
        continue;
    }

    // something else, but not an integer is there ....
    // if you cannot recover the lopop will exit 
}