Question

I am writing a UTF-8 library for C++ as an exercise as this is my first real-world C++ code. So far, I’ve implemented concatenation, character indexing, parsing and encoding UTF-8 in a class called “ustring”. It looks like it’s working, but two seemingly equivalent ways of declaring a new ustring behave differently. The first way:

ustring a;
a = "test";

works, and the overloaded “=” operator parses the string into the class (which stores the Unicode strings as an dynamically allocated int pointer). However, the following does not work:

ustring a = "test";

because I get the following error:

test.cpp:4: error: conversion from ‘const char [5]’ to non-scalar type ‘ustring’ requested

Is there a way to workaround this error? It probably is a problem with my code, though. The following is what I’ve written so far for the library:

#include <cstdlib>
#include <cstring>
class ustring {
  int * values;
  long len;
  public:
  long length() {
    return len;
  }
  ustring * operator=(ustring input) {
    len = input.len;
    values = (int *) malloc(sizeof(int) * len);
    for (long i = 0; i < len; i++)
      values[i] = input.values[i];
    return this;
  }
  ustring * operator=(char input[]) {
    len = sizeof(input);
    values = (int *) malloc(0);
    long s = 0;                                                                 // s = number of parsed chars
    int a, b, c, d, contNeed = 0, cont = 0;
    for (long i = 0; i < sizeof(input); i++)
      if (input[i] < 0x80) {                                                    // ASCII, direct copy (00-7f)
        values = (int *) realloc(values, sizeof(int) * ++s);
        values[s - 1] = input[i];
      } else if (input[i] < 0xc0) {                                             // this is a continuation (80-bf)
        if (cont == contNeed) {                                                 // no need for continuation, use U+fffd
          values = (int *) realloc(values, sizeof(int) * ++s);
          values[s - 1] = 0xfffd;
        }
        cont = cont + 1;
        values[s - 1] = values[s - 1] | ((input[i] & 0x3f) << ((contNeed - cont) * 6));
        if (cont == contNeed) cont = contNeed = 0;
      } else if (input[i] < 0xc2) {                                             // invalid byte, use U+fffd (c0-c1)
        values = (int *) realloc(values, sizeof(int) * ++s);
        values[s - 1] = 0xfffd;
      } else if (input[i] < 0xe0) {                                             // start of 2-byte sequence (c2-df)
        contNeed = 1;
        values = (int *) realloc(values, sizeof(int) * ++s);
        values[s - 1] = (input[i] & 0x1f) << 6;
      } else if (input[i] < 0xf0) {                                             // start of 3-byte sequence (e0-ef)
        contNeed = 2;
        values = (int *) realloc(values, sizeof(int) * ++s);
        values[s - 1] = (input[i] & 0x0f) << 12;
      } else if (input[i] < 0xf5) {                                             // start of 4-byte sequence (f0-f4)
        contNeed = 3;
        values = (int *) realloc(values, sizeof(int) * ++s);
        values[s - 1] = (input[i] & 0x07) << 18;
      } else {                                                                  // restricted or invalid (f5-ff)
        values = (int *) realloc(values, sizeof(int) * ++s);
        values[s - 1] = 0xfffd;
      }
    return this;
  }
  ustring operator+(ustring input) {
    ustring result;
    result.len = len + input.len;
    result.values = (int *) malloc(sizeof(int) * result.len);
    for (long i = 0; i < len; i++)
      result.values[i] = values[i];
    for (long i = 0; i < input.len; i++)
      result.values[i + len] = input.values[i];
    return result;
  }
  ustring operator[](long index) {
    ustring result;
    result.len = 1;
    result.values = (int *) malloc(sizeof(int));
    result.values[0] = values[index];
    return result;
  }
  char * encode() {
    char * r = (char *) malloc(0);
    long s = 0;
    for (long i = 0; i < len; i++) {
      if (values[i] < 0x80)
        r = (char *) realloc(r, s + 1),
        r[s + 0] = char(values[i]),
        s += 1;
      else if (values[i] < 0x800)
        r = (char *) realloc(r, s + 2),
        r[s + 0] = char(values[i] >> 6 | 0x60),
        r[s + 1] = char(values[i] & 0x3f | 0x80),
        s += 2;
      else if (values[i] < 0x10000)
        r = (char *) realloc(r, s + 3),
        r[s + 0] = char(values[i] >> 12 | 0xe0),
        r[s + 1] = char(values[i] >> 6 & 0x3f | 0x80),
        r[s + 2] = char(values[i] & 0x3f | 0x80),
        s += 3;
      else
        r = (char *) realloc(r, s + 4),
        r[s + 0] = char(values[i] >> 18 | 0xf0),
        r[s + 1] = char(values[i] >> 12 & 0x3f | 0x80),
        r[s + 2] = char(values[i] >> 6 & 0x3f | 0x80),
        r[s + 3] = char(values[i] & 0x3f | 0x80),
        s += 4;
    }
    return r;
  }
};

Answer 1

Your problem is that ustring a = "test" actually invokes the constructor, not the assignment operator. yay, welcome to c++ 🙂

You’ll need to define yourself both a default constructor and one that takes a const char*, because once you define a constructor, you need to define all your constructors.

A few other things:

• pass your input ustring by reference
• pass const char * instead of char[] (you don’t modify the input and char* is more common)
• sizeof isn’t doing what you think it’s doing, it doesn’t work properly for array parameters. It is returning you sizeof(char*), not sizeof(array).
• return reference to this from your operators.
• you can use vector<int> values; to manage all your memory for you.
• encode() should probably return a string. With string:
  • it manages its own memory, so the caller doesn’t need to free or delete it.
  • you can use s.append(c); instead of using realloc.
  • you can use printf("%s", s.c_str());, but in c++ you usually use cout << s;
• consider defining a copy constructor as well.

Like this:

class ustring {
 public:
  // Default constructor, allows you to create your class with no arguments.
  ustring() { ...; }
  // Allows you to create your class from string literals.
  ustring(const char *input) { ...; }
  // Copy constructor, allows you to create your class from other instances.
  ustring(const ustring &input) { ...; }

  // Assignment operators.
  ustring &operator=(const ustring &input) { ...; return *this; }
  ustring &operator=(const char *input) { ...; return *this; }
};

int main() {
  ustring s, t;  // invokes default constructor.
  s = t;         // invokes ustring assignment op.
  s = "test";    // invokes const char* assignment op.
  ustring u = "test";  // invokes const char* constructor.
  ustring v("test");   // invokes const char* constructor.
  ustring x(u);  // invokes copy constructor.
}

If this is c++, why are you doing all this malloc/realloc stuff? I haven’t fully parsed that code, but I’d imagine there’s a simpler way… see the comment about using vector.

As @Michael Aaron Safyan mentioned in the comments, if you do any memory allocation for the ustring class, you will want to deallocate it in the destructor. However, I think by switching to memory managed containers – vector & string – you’ll avoid any of your own memory management and can avoid writing a destructor.