从字符串中删除重复字符

Question

I have a string, like eg acaddef or bbaaddgg . I have to remove from it, as fast as possible, all repeating characters. So, for example, pooaatat after should look like poat and ggaatpop should look like gatpo . Is there any built-in function or algorithm to do that quickly? I tried to search STL, but without satisfaing result.

Answer 1

Okay, so here are 4 different solutions.

Fixed Array

std::string str = "pooaatat";

// Prints "poat"
short count[256] = {0};
std::copy_if(str.begin(), str.end(), std::ostream_iterator<char>(std::cout),
             [&](unsigned char c) { return count[c]++ == 0; });

Count Algorithm + Iterator

std::string str = "pooaatat";

// Prints "poat"
std::string::iterator iter = str.begin();
std::copy_if(str.begin(), str.end(), std::ostream_iterator<char>(std::cout),
             [&](char c) { return !std::count(str.begin(), iter++, c); });

Unordered Set

std::string str = "pooaatat";

// Prints "poat"
std::unordered_set<char> container;
std::copy_if(str.begin(), str.end(), std::ostream_iterator<char>(std::cout),
             [&](char c) { return container.insert(c).second; });

Unordered Map

std::string str = "pooaatat";

// Prints "poat"
std::unordered_map<char, int> container;
std::copy_if(str.begin(), str.end(), std::ostream_iterator<char>(std::cout),
             [&](char c) { return container[c]++ == 0; });

Answer 2

AFAIK, there is no built-in algorithm for doing this. The std::unique algorithm is valid if you want to remove only consecutive duplicate characters.

However you can follow the following simple approach:

If the string contains only ASCII characters, you can form a boolean array A[256] denoting whether the respective character has been encountered already or not.

Then simply traverse the input string and copy the character to output if A[character] is still 0 (and make A[character] = 1).

In case the string contains arbitrary characters, then you can use a std::unordered_map or a std::map of char to int.

Answer 3

Built-in regular expressions should be efficient, ie

#include <regex>
[...]

const std::regex pattern("([\\w ])(?!\\1)");
string s = "ssha3akjssss42jj 234444 203488842882387 heeelloooo";
std::string result;

for (std::sregex_iterator i(s.begin(), s.end(), pattern), end; i != end; ++i)
    result.append((*i)[1]);

std::cout << result << std::endl;

Of course, you can modify the cpaturing group to your needs. The good thing is that it is supported in Visual Studio 2010 tr1 already. gcc 4.8, however, seems to have a problem with regex iterators.