Add dataset parse/load library

[wzhao18]

No description provided.

[anandj91]

struct Row {
    string dataset;
};

struct CSVRow : public Row {
    vector<string> cols;
};

template<typename R>
class DataParser {
protected:
    virtual void run() = 0;
    virtual void parse(fstream&) = 0;
    virtual void decode(R&, stream::stream_event&) = 0;

    bool write_serialized_to_ostream(stream::stream_event &t)
    {
        if (!google::protobuf::util::SerializeDelimitedToOstream(t, &cout)) {
            cerr << "Fail to serialize data into output stream" << endl;
            return false;
        }
        return true;
    }

};

class CSVParser : public DataParser<CSVRow> {
protected:
    void parse_file(fstream &file)
    {   
        string line;
        getline(file, line);

        while (getline(file, line)) {
            CSVRow row;
            string word;
            stringstream ss(line);

            while (getline(ss, word, ',')) {
                row.cols.push_back(word);
            }

            stream::stream_event data;
            decode(row, data);
            if (!write_serialized_to_ostream(data)) {
                break;
            }
        }
    }   
};

consider something like the above organization of the base classes.

Additionally, try to follow some standard styling practice when you write code. Like the position of braces of classes and functions.

[anandj91]

Consider using two different parsers for fare and trip datasets. You don't need to combine them in one.

[wzhao18]

The intention of combining the parsers for fare and trip datasets is that both can be streamed to std out at the same time. The taxi benchmark requires two streams.