C++ - 如何使用 stream 解析文件？

Question

I have a file and I need to loop through it assigning an int foo, string type, 64/128 bit long. How would I use a stream to parse these lines into the following variables - I want to stick with the stream syntax ( ifs >> foo >> type ) but in this case type would end up being the rest of the line after the 0/52... and at that point I'd just get a char* and use strtoull and such so why use the stream in the first place... I'm hoping for readable code without horrid performance over char strings / strtok / strtoull

//input file:
0ULL'04001C0180000000000000000EE317BC'
52L'04001C0180000000'
//ouput:
//0 ULL 0x04001C0180000000 0x000000000EE317BC
//52 L 0x04001C0180000000

  ifstream ifs("input.data");
  int foo;
  string type;
  unsigned long long ull[2];

Answer 1

Boost Spirit implementation

Here is the mandatory Boost Spirit (Qi) based implementation. For good measure, including formatting using Boost Spirit (Karma):

#include <string>
#include <iostream>
#include <fstream>
#include <boost/spirit/include/qi.hpp>
#include <boost/spirit/include/karma.hpp>

namespace karma=boost::spirit::karma;
namespace qi   =boost::spirit::qi;

static qi::uint_parser<unsigned long long, 16, 16, 16> hex16_p; // parse long hex
static karma::uint_generator<unsigned long long, 16>   hex16_f; // format long hex

int main(int argc, char** args)
{
    std::ifstream ifs("input.data");
    std::string line;
    while (std::getline(ifs, line))
    {
        std::string::iterator begin = line.begin(), end = line.end();

        int                             f0;
        std::string                     f1;
        std::vector<unsigned long long> f2;

        bool ok = parse(begin, end,
                qi::int_                    // an integer
                >> *qi::alpha               // alternatively: *(qi::char_ - '\'')
                >> '\'' >> +hex16_p >> '\'' // accepts 'n x 16' hex digits
            , f0, f1, f2);

        if (ok)
            std::cout << "Parsed: " << karma::format(
                 karma::int_ 
                 << ' ' << karma::string 
                 << ' ' << ("0x" << hex16_f) % ' '
             , f0, f1, f2) << std::endl;
        else
            std::cerr << "Parse failed: " << line << std::endl;
    }

    return 0;
}

Test run:

Parsed: 0 ULL 0x4001c0180000000 0xee317bc
Parsed: 52 L 0x4001c0180000000

^{see Tweaks and samples below for info on how to tweak eg hex output}

Benchmark

I had benchmarked @Cubbi's version and the above as written on 100,000x the sample inputs you provided. This initially gave Cubbi's version a slight advantage: 0.786s versus 0.823s .

Now, that of course wasn't fair comparison, since my code is constructing the parser on the fly each time. With that taken out of the loop like so:

typedef std::string::iterator It;

const static qi::rule<It> parser = qi::int_ >> *qi::alpha >> '\'' >> +hex16_p >> '\'';
bool ok = parse(begin, end, parser, f0, f1, f2);

Boost Spirit comes out a clear winner with only 0.093s ; already a factor 8.5x faster, and that is even with the karma formatter still being constructed each iteration.

^{with the output formatting commented out in both versions, Boost Spirit is >11x faster}

Tweaks, samples

Note how you can easily tweak things:

//  >> '\'' >> +hex16_p >> '\'' // accepts 'n x 16' hex digits
    >> '\'' >> qi::repeat(1,2)[ hex16_p ] >> '\'' // accept 16 or 32 digits

Or format the hex output just like the input:

// ("0x" << hex16_f) % ' '
karma::right_align(16, '0')[ karma::upper [ hex16_f ] ] % ""

Changed sample output:

0ULL'04001C0180000000000000000EE317BC'
Parsed: 0 ULL 04001C0180000000000000000EE317BC
52L'04001C0180000000'
Parsed: 52 L 04001C0180000000

HTH

Answer 2

This is a rather trivial task for a more sophisticated parser such as boost.spirit .

To solve this using just the standard C++ streams you would need to

• a) treat ' as whitespace and
• b) take an extra pass over the string "04001C0180000000000000000EE317BC" which has no separators between the values.

Borrowing Jerry Coffin's sample facet code ,

#include <iostream>
#include <fstream>
#include <locale>
#include <vector>
#include <sstream>
#include <iomanip>
struct tick_is_space : std::ctype<char> {
    tick_is_space() : std::ctype<char>(get_table()) {}
    static std::ctype_base::mask const* get_table()
    {
        static std::vector<std::ctype_base::mask>
               rc(table_size, std::ctype_base::mask());
        rc['\n'] = std::ctype_base::space;
        rc['\''] = std::ctype_base::space;
        return &rc[0];
    }
};

int main()
{
    std::ifstream ifs("input.data");
    ifs.imbue(std::locale(std::locale(), new tick_is_space()));
    int foo;
    std::string type, ullstr;
    while( ifs >> foo >> type >> ullstr)
    {
        std::vector<unsigned long long> ull;
        while(ullstr.size() >= 16) // sizeof(unsigned long long)*2
        {
            std::istringstream is(ullstr.substr(0, 16));
            unsigned long long tmp;
            is >> std::hex >> tmp;
            ull.push_back(tmp);
            ullstr.erase(0, 16);
        }
        std::cout << std::dec << foo << " " << type << " "
                  << std::hex << std::showbase;
        for(size_t p=0; p<ull.size(); ++p)
            std::cout << std::setw(16) << std::setfill('0') << ull[p] << ' ';
        std::cout << '\n';
    }
}

test: https://ideone.com/lRBTq