fast way to parse a configuration

Question

Say you have

char *=  "name:454";

What is the best way to parse name and the number, thus

std:string id would equal to "name";

double d would equal to 454;

STL please, no boost.

Answer 1

You want to look at the strtok function using ':' as the token. Here is an example:

#include <stdio.h>
#include <string.h>

int main ()
{
    char str[] = "name:454";
    char* c = strtok(str, ":");

    while (c != NULL)
    {
     printf("%s\n", c);
     c = strtok(NULL, ":");
    }

    return 0;
}

Answer 2

I'd use getline.

Example: http://github.com/lennartkoopmann/scopeport-server/blob/23e3d2e2fede7ced0810deee0aa501a70a4eba40/src/core.cc#L1353

Answer 3

Not the stl solution you asked for; however, sooner or later you'll need to parse more complicated expressions (hint - sooner than you expect), at which point you will need regular expressions - so why not start now. Your expression is captures by:

^(\w+):(\d+)$

I'm no boost fanatic, but their regexp library is nice.

Answer 4

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

int main() {
  /* output storage */
  std::string id; 
  double d;
  /* convert input to a STL string */
  std::string s("name:454");
  size_t off = std::string::npos;
  /* smart replace: parsing is easier with a space */    
  if ((off = s.find(':')) != std::string::npos) { // error check: all or none 
    s = s.replace(off, 1, 1, ' ');
    std::istringstream iss(s);
    iss >> id >> d;
    std::cout << "id = " << id << " ; d = " << d << '\n';
  }
  return 0;
}

Though, I'd just write my own parser or use the C scanner function for speed.

Answer 5

I would write my own. The basic idea would be to read one character at a time from the stream. If it isn't a colon, append it to the id string. If it is, skip it, then use the istream >> operator to load an integer, double float, or whatever is needed. Then you probably put the result into a std::map.

Answer 6

Another possibility is to use an already written parser for a simple format like INI.

Here is one: http://code.jellycan.com/SimpleIni/

I looked at the code for SimpleIni and it isn't very C++ and STL'ish, but do you really care?

Answer 7

template < typename T >
T valueFromString( const std::string &src )
{
    std::stringstream s( src );
    T result = T();
    s >> result;
    return result;
}

std::string wordByNumber( const std::string &src, size_t n, const std::string &delims )
{
    std::string::size_type word_begin = 0;
    for ( size_t i = 0; i < n; ++i )
    {
        word_begin = src.find_first_of( delims, word_begin );
    }
    std::string::size_type word_end = src.find_first_of( delims, word_begin );

    word_begin = std::string::npos == word_begin || word_begin == src.length() ? 0 : word_begin + 1;
    word_end = std::string::npos == word_end ? src.length() : word_end;
    return src.substr( word_begin, word_end - word_begin);
}


char *a = "asdfsd:5.2";
std::cout << wordByNumber( a, 0, ":" ) << ", " << valueFromString< double > ( wordByNumber( a, 1, ":" ) );

PS: in previous revision I've posted wordByNumber - which skipped neighbored delimiters (e.g.:::), in current revision they are treated as empty word.