Program crashes - bad_alloc when creating new char array

Question

I have a C++ function that splits a char array into multiple char arrays when it encounters a delimiter. For some reason, when saving the third split array the program just crashes and sometimes returns an std::bad_alloc exception.

char ** explode(const char * arr, const char delim) {
int start, end, curr=0, count=1;
char ** strings;
//Iegūst explodēto stringu skaitu
for (int i = 0; arr[i] != 0; i++) {
    if (arr[i] == delim && i != 0 && arr[i+1] != 0 && arr[i+1] != delim ) { //Nav pirmais, nav pēdējais, nav pa labi vēlviens delimiters
        count++;
    }
}
strings = new char*[count];
start = 0;
for (int i = 0; arr[i] != 0; i++) {
    if (arr[i] == delim || arr[i+1] == 0) {
        if (arr[i] == delim) {
            end = i;
        } else {
            end = i+1;
        }
        if (end-start < 1) {
            start++;
        } else {
            copystring(arr,strings[curr++],start,end-start);
            start = i+1;
        }
    }
}
for (int i = 0; i < count; i++) {
    cout << strings[i] << endl;
}

return strings;
}

//Pārkopē daļu no pirmā char masīva uz otru, no START pozīcijas, līdz GARUMS garumā
void copystring(const char * from, char *& to, const int start, const int garums) {
    int curr=0;
    if (garums < 1 || start > charlen(from)) {
        return;
    }
    to = new char[garums];
    for (int i = start; i < start+garums && from[i] != 0; i++) {
        to[curr++] = from[i];
    }
    to[curr] = 0;
}

It's hard to tell because it doesn't really tell me at which line everything goes wrong, but I think it happens at

to = new char[garums];

I've tried debugging this line within CodeBlocks, but for some reason when using breakpoints and tracking the variables the applications works fine and executes correctly. It only crashes when running it normally, without debugging...

Also note, that I can't use strings or pretty much any library except fstream and iostream.

EDIT: I tried changing the new char[garums] part to new char[100] and it magically started working. The problem is that I then changed it to new char[10] in which case everything still worked. I even outputted the saved text to the console and it saved everything properly. How could it have saved big words in a char array that is 10 character long (the words I'm testing are longer than 10 characters)? When I changed it to new char[1] however it started crashing again, but again only after the 3rd loop iteration. So it somehow saved the first 2 words in a 1 character long array?

EDIT2: And now it magically started working even with new char[garums]. Something is really wrong here, anyone have any ideas?

Answer 1

Since I don't know what input data you have I will have to guess:

Here you allocate your pointer array but please note that all the pointers are uninitialized.

strings = new char*[count]

then when you parse the code you use a variable curr which you let run freely so it is not certain that all strings[] have been set to some value or whether curr lands on a number larger than count.

If I were you I would put in a check to make sure that:

a) curr does not exceed count

b) that if curr lands on a value less than count, set the rest of the pointers to nullptr

Answer 2

The bug you refer to in your question likely crops up when trying to use the pointer to pointer being returned from the explode function. Some pointers ; If you have to write C code, don't use a mishmash of C/C++, Use the library functions rather than re-inventing the wheel (strncpy in copystring) Your word count was off because you didn't take into account the word between the last delimiter and EOL Below are some minor changes to your code as a complete example :

#include <stdio.h>
#include <strings.h>

void copystring(const char *from, char **to, const int numchars)
{
    if (numchars > 0) {
            *to = new char[numchars];
            strncpy(*to, from, numchars) ;
            (*to)[numchars] = '\0' ;
    }
}

char **explode(const char * buffer, const char delim)
{
    int count = 0 ;

    if (strlen(buffer) > 0) {
            int inword = 0 ;
            int idx = 0 ;
            do {
                    if (buffer[idx] == delim || buffer[idx] == '\0') {
                            if (inword == 1) {
                                    count++ ;
                                    inword = 0 ;
                            }
                    } else {
                            inword = 1 ;
                    }
            } while (buffer[idx++] != 0) ;
    }

    int start = 0;
    int end = 0 ;
    int curr = 0 ;
    int idx = 0 ;

    char **values = new char*[count+1];

    do {
            if (buffer[idx] == delim || buffer[idx] == '\0') {
                    end = idx;
                    if (end-start > 0) {
                            copystring(&buffer[start], &values[curr++], end - start) ;
                    }
                    start = ++end ;
            }
    } while (buffer[idx++] != 0) ;

    values[curr] = NULL ;
    for (int idx = 0; idx < count; idx++) {
            fprintf(stdout, "'%s'\n", values[idx]) ;
    }

    return values;
}

int main(int argc, char *argv[])
{
    char inputstr[] = "The, quick, brown, fox, jumped, over,,, ,,, ,,,,, ,,,, the, lazy, dog's, back" ;
    char **values = explode(inputstr, ',') ;

    while (*values != NULL) {
            fprintf(stdout, "%s\n" , *values) ;
            *values++ ;
    }

    return (0) ;
}

Answer 3

This has probably to do with to being of type char*& instead of type char*. On the other hand, I never programmed C++ like this (are you sure that this is not C?). Using explicit memory management (like ´new´) is as good as playing playing russian roulette.

Here is a more standard C++ way of doing this:

#include <vector>
#include <string>
#include <iostream>

std::vector<std::string> splitString(std::string& str, char c) {
  std::vector<std::string> substrings;
  while(true) {
    unsigned pos = str.find(c);
    substrings.push_back(str.substr(0,pos));
    if(pos == std::string::npos) break;
    str = str.substr(pos+1);
  }
  return substrings;
}

int main()
{
  char c = '*';
  std::string str = "Some*string*that we need to split*properly*";
  std::vector<std::string> result = splitString(str,c);

  for(unsigned i = 0; i < result.size(); ++i) {
    std::cout << i << ": " << result[i] << "\n";
  }
}

Output:

0: Some
1: string 
2: that we need to split
3: properly
4: