Splitting a string with strtok_s

Question

I am trying to split a string by a specified delimiter according to this example: http://msdn.microsoft.com/en-us/library/ftsafwz3(v=VS.90).aspx

My code compiles without errors in Visual C++ 2010, but when I want to run it, I get this error message:

Unhandled exception at 0x773a15de in Test.exe: 0xC0000005: Access violation reading location 0x00000000.

Here's my code:

#include "stdafx.h"
#include <iostream>
#include <fstream>
#include <string>
#include <sstream>
#include <regex>

using namespace std;

vector<char *> splitString(char in[])
{
vector<char *> parts(15);
char seps[]   = " ,\t\n";
char *next_token1 = NULL;
char *token1 = NULL;
token1 = strtok_s(in, seps, &next_token1);
while ((token1 != NULL))
{
    if (token1 != NULL)
    {
        token1 = strtok_s( NULL, seps, &next_token1);
                    //printf( " %s\n", token1 );
        parts.push_back(token1);
    }
}
return parts;
}

int main(int argc, char * argv[])
{
char string1[] =
    "A string\tof ,,tokens\nand some  more tokens";
vector<char *> parts=splitString(string1);
cout << parts[0] <<endl;
cout << parts[1] <<endl;
return 0;
}

It seems to be illegal that I try to display the vector's elements, but why?

The vector's capacity should be sufficient and a

printf( " %s\n", token1 );

in the while loop prints out the tokens!

Answer 1

The use of vector is incorrect. It is constructed with 15 elements and then push_back() is used to add the strings which adds new elements after the initial 15. This means the first 15 elements are unassigned (null):

std::cout << parts[0] << end; // parts[0] is null

Either:

• don't preallocate elements at construction, or
• use operator[] and not push_back() (add additional loop terminator to protect going beyond the end of the vector)

(Consider changing to std::vector<std::string>.)

Just to mention boost::split() that can produces a list of tokens (std::vector<std::string>) from an input string and permits specification of multiple delimiters.

Answer 2

The trouble is A general word of caution: you're using strtok (family of) functions in c++. Note that this old API modifies it's argument. This is frequently not what you expect, and for this reason I'd advise against using this C library function.

Furthermore you are assuming that 15 elements will be read, leaving the 'surplus' elements uninitialized. This, too, results in undefined behaviour on accessing those elements.

---

May I suggest a C++ approach, since you are using it:

#include <iostream>
#include <sstream>
#include <iterator>
#include <algorithm>

using namespace std;

vector<std::string> splitString(const char in[])
{
    std::istringstream iss(in);
    std::istream_iterator<std::string> first(iss), last;

    std::vector<std::string> parts;
    std::copy(first, last, std::back_inserter(parts));
    return parts;
}

int main(int argc, char * argv[])
{
    const char string1[] = "A string\tof ,,tokens\nand some  more tokens";
    vector<std::string> parts = splitString(string1);
    cout << parts[0] <<endl;
    cout << parts[1] <<endl;
    return 0;
}

This uses the fact that, by default, iostreams will skipws (skip whitespace)