sort std::list case sensitive elements

Question

#include <list>
#include <string>
using std::string;
using std::list;

int main()
{
    list <string> list_;
    list_.push_back("C");
    list_.push_back("a");
    list_.push_back("b");

    list_.sort();
}

does sort() function sort the elements according to their character codes? I want the result here to be a b C after the sorting is done.

Answer 1

The default comparator (<) using the default char_traits< char > will sort your list as C a b.

See list::sort.

In order to achieve the desired order a b C you can either:

1. compose your list of string types with custom char_traits, or

2. provide an instance of a custom string comparator to sort, e.g.

   bool istring_less(const string& lhs, const string& rhs) {
     string::const_iterator \
       lb = lhs.begin(), le = lhs.end(),
       rb = rhs.begin(), re = rhs.end();
     const char lc, rc;
     for ( ; lb != le && rb != re; ++lb, ++rb) {
       lc = tolower(*lb);
       rc = tolower(*rb);
       if (*lc < *rc) return true;
       if (*lc > *rc) return false;
     }
     // if rhs is longer than lhs then lhs<rhs
     return (rb != re);
   }
   ...
   list.sort(istring_less);
   

Answer 2

yes

you can check the following code to use a custom comparator

Answer 3

Case-insensitive character comparisons are tricky if you want to support characters from other languages. That's why it's a good idea to do them in a locale-sensible manner:

struct char_iless 
: public std::binary_function<char, char, bool>
{
    std::locale loc;

    char_iless(std::locale const & loc=std::locale()) : loc(loc) 
    {
    }

    bool operator()(char a, char b) const
    {
        return std::tolower(a, loc) < std::tolower(b, loc);
    }
};

This is how you use this class to compare two chars:

char_iless('a', 'b', my_locale);

Just use std::locale() as my_locale if you want to use the one that's set as default.

If you can use Boost then there is am is_iless functor in the String Algorithms library which does the same thing.

Extending this from comparing chars to strings is easy thanks to std::lexicographical_compare:

struct str_iless 
: public std::binary_function<std::string, std::string, bool>
{
    std::locale loc;

    str_iless(std::locale const & loc=std::locale()) : loc(loc) 
    {
    }

    bool operator()(std::string const & a, std::string const & b) const
    {
        return std::lexicographical_compare(
            a.begin(), a.end(),
            b.begin(), b.end(),  
            char_iless(loc)
        );
    }
};

Now you have all that it's required to solve your problem:

int main()
{
    std::list<std::string> list;
    list.push_back("C");
    list.push_back("a");
    list.push_back("b");

    // Sort using default locale
    list.sort(str_iless());  

    // Sort using French locale 
    // (warning: this locale format string is MS specific)
    std::locale loc("French_France.1252");
    list.sort(str_iless(loc));
}

Answer 4

Here are what I consider to be some cleaner and one significantly faster alternative:

#include    <string>
#include    <cstring>
#include    <iostream>
#include    <boost/algorithm/string.hpp>

using std::string;
using std::list;
using std::cout;
using std::endl;

using namespace boost::algorithm;

// recommended in Meyers, Effective STL when internationalization and embedded
// NULLs aren't an issue.  Much faster than the STL or Boost lex versions.
struct ciLessLibC : public std::binary_function<string, string, bool> {
    bool operator()(const string &lhs, const string &rhs) const {
        return strcasecmp(lhs.c_str(), rhs.c_str()) < 0 ? 1 : 0;
    }
};

// If you need sorting according to the default system local
struct ciLessBoost : std::binary_function<std::string, std::string, bool>
{
    bool operator() (const std::string & s1, const std::string & s2) const {
        return lexicographical_compare(s1, s2, is_iless());
    }
};

int main(void) {
    list <string> list_;
    list_.push_back("C");
    list_.push_back("a");
    list_.push_back("b");

    list_.sort(ciLessLibC());
    list_.sort(ciLessBoost());

    return 0;
}