在自定义迭代器上应用reverse_iterator后引用无效

Reference invalidation after applying reverse_iterator on a custom made iterator

本文关键字:iterator 引用 无效 reverse 自定义 迭代器 应用      更新时间:2023-10-16

我实现了一个双向迭代器,但它不是对数据结构进行操作,而是返回一个数学级数,可以在两个方向上迭代计算。事实上,我正在对整数进行迭代,在int上使用++和--。这意味着数据不会存储在不同的结构中,因此当迭代器超出范围时,值也会超出范围。

尽管如此,我还是希望下一个代码(最小失败示例)示例能够工作,因为迭代器始终处于作用域中。但它不起作用:(

#include <iostream>
#include <iterator>
#include <vector>
class my_iterator : public std::iterator<std::bidirectional_iterator_tag, int> {
  int d_val = 12;
public:
  my_iterator  operator--(int) { std::cout << "decrement--n"; return my_iterator(); }
  my_iterator &operator--()    { std::cout << "--decrementn"; return *this; }
  my_iterator  operator++(int) { std::cout << "increment++n"; return my_iterator(); }
  my_iterator &operator++()    { std::cout << "++incrementn"; return *this; }
  int &operator*() { std::cout << "*dereferencen"; return d_val; }
  bool operator==(my_iterator const  &o) { return false; }
  bool operator!=(my_iterator const  &o) { return true ; }
};

int main() {
  auto it = std::reverse_iterator<my_iterator>();
  int &i = *it;
  if (true)
  {
    std::cout << i << 'n';
  }
  else
  {
    std::vector<int> vec;
    vec.push_back(i);
    std::cout << vec[0] << 'n';
  }
}

来源:http://ideone.com/YJKvpl

if分支导致内存冲突,正如valgrind:正确检测到的那样

--decrement
*dereference
==7914== Use of uninitialised value of size 8
==7914==    at 0x4EC15C3: ??? (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.20)
==7914==    by 0x4EC16FB: std::ostreambuf_iterator<char, std::char_traits<char> > std::num_put<char, std::ostreambuf_iterator<char, std::char_traits<char> > >::_M_insert_int<long>(std::ostreambuf_iterator<char, std::char_traits<char> >, std::ios_base&, char, long) const (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.20)
==7914==    by 0x4EC1C7C: std::num_put<char, std::ostreambuf_iterator<char, std::char_traits<char> > >::do_put(std::ostreambuf_iterator<char, std::char_traits<char> >, std::ios_base&, char, long) const (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.20)
==7914==    by 0x4ECEFB9: std::ostream& std::ostream::_M_insert<long>(long) (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.20)
==7914==    by 0x40087B: main (divine.cc:25)
==7914== 
==7914== Conditional jump or move depends on uninitialised value(s)
==7914==    at 0x4EC15CF: ??? (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.20)
==7914==    by 0x4EC16FB: std::ostreambuf_iterator<char, std::char_traits<char> > std::num_put<char, std::ostreambuf_iterator<char, std::char_traits<char> > >::_M_insert_int<long>(std::ostreambuf_iterator<char, std::char_traits<char> >, std::ios_base&, char, long) const (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.20)
==7914==    by 0x4EC1C7C: std::num_put<char, std::ostreambuf_iterator<char, std::char_traits<char> > >::do_put(std::ostreambuf_iterator<char, std::char_traits<char> >, std::ios_base&, char, long) const (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.20)
==7914==    by 0x4ECEFB9: std::ostream& std::ostream::_M_insert<long>(long) (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.20)
==7914==    by 0x40087B: main (divine.cc:25)
==7914== 
==7914== Conditional jump or move depends on uninitialised value(s)
==7914==    at 0x4EC1724: std::ostreambuf_iterator<char, std::char_traits<char> > std::num_put<char, std::ostreambuf_iterator<char, std::char_traits<char> > >::_M_insert_int<long>(std::ostreambuf_iterator<char, std::char_traits<char> >, std::ios_base&, char, long) const (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.20)
==7914==    by 0x4EC1C7C: std::num_put<char, std::ostreambuf_iterator<char, std::char_traits<char> > >::do_put(std::ostreambuf_iterator<char, std::char_traits<char> >, std::ios_base&, char, long) const (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.20)
==7914==    by 0x4ECEFB9: std::ostream& std::ostream::_M_insert<long>(long) (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.20)
==7914==    by 0x40087B: main (divine.cc:25)
==7914== 
12
==7914== 
==7914== HEAP SUMMARY:
==7914==     in use at exit: 0 bytes in 0 blocks
==7914==   total heap usage: 0 allocs, 0 frees, 0 bytes allocated
==7914== 
==7914== All heap blocks were freed -- no leaks are possible
==7914== 
==7914== For counts of detected and suppressed errors, rerun with: -v
==7914== Use --track-origins=yes to see where uninitialised values come from
==7914== ERROR SUMMARY: 5 errors from 3 contexts (suppressed: 0 from 0)

其他分支不会导致内存冲突,或者至少在我的valgrind所能检测到的范围内。然而,存储在矢量中的值是"随机的":

--decrement
*dereference
-16777520

我对发生的事情有点惊讶。迭代器应该一直在作用域中,但引用似乎已失效。为什么在打印12时我会收到内存违规,或者为什么在存储与12不同的内容时我不会收到这些违规?

reverse_iterator不适用于所谓的"隐藏迭代器",即返回对自身内容的引用的迭代器。reverse_iteratoroperator*生成包装迭代器的副本,对其进行递减,并返回对副本的解引用结果。因此,如果取消引用迭代器返回对其内部某个内容的引用,则该引用将变为悬空。

在C++11规范中曾试图使其工作,但事实证明,如果不为非隐藏迭代器添加大量开销*,就不可能实现,因此该规范被恢复到C++03版本。


*要支持"隐藏迭代器",必须添加一个额外的数据成员来存储递减的当前迭代器,使reverse_iterator的大小加倍;然后必须使用某种形式的同步,因为CCD_ 5是CCD_。对于这样一个不常见的用例,要向所有reverse_iterator添加大量开销

如前所述,C++03和C++14标准以这种方式定义reverse_iterator::operator*

24.5.1.3.4 operator*[反向iter.op.star]

reference operator*() const;

1效果:

Iterator tmp = current;
return *--tmp;

operator*返回后,tmp将被销毁,因此对存储在tmp内部的数据的任何引用都将无效。C++11标准改变了这一点,并添加了一个注释:

24.5.1.3.4 operator*[反向iter.op.star]

reference operator*() const;

1效果:

deref_tmp = current;
--deref_tmp;
return *deref_tmp;

2[注意:此操作必须使用辅助成员变量,而不是一个临时变量,用于避免返回持续超过其关联迭代器的生存期。(见24.2.)——尾注]

这实际上是不可能实现的,因为operator*上有const限定符,所以在C++11和C++14之间恢复了措辞。

最好的解决方案可能是基于针对任何版本的C++的C++11措辞来实现您自己的reverse_iterator版本。幸运的是,该规范非常简单易懂。作为一个工作示例,下面是我为C++14编写的一个示例:

template <class Iterator>
class stashing_reverse_iterator :
  public std::iterator<
    typename std::iterator_traits<Iterator>::iterator_category,
    typename std::iterator_traits<Iterator>::value_type,
    typename std::iterator_traits<Iterator>::difference_type,
    typename std::iterator_traits<Iterator>::pointer,
    typename std::iterator_traits<Iterator>::reference
  > {
  typedef std::iterator_traits<Iterator> traits_type;
public:
  typedef Iterator iterator_type;
  typedef typename traits_type::difference_type difference_type;
  typedef typename traits_type::reference       reference;
  typedef typename traits_type::pointer         pointer;
  stashing_reverse_iterator() : current() {}
  explicit stashing_reverse_iterator(Iterator x) : current(x) {}
  template <class U>
  stashing_reverse_iterator(const stashing_reverse_iterator<U>& u) : current(u.current) {}
  template <class U>
  stashing_reverse_iterator& operator=(const stashing_reverse_iterator<U>& u) {
    current = u.base();
    return *this;
  }
  Iterator base() const {
    return current;
  }
  // Differs from reverse_iterator::operator*:
  // 1. const qualifier removed
  // 2. current iterator is stored in a member field to ensure references are
  //    always valid after this function returns
  reference operator*() {
    deref_tmp = current;
    --deref_tmp;
    return *deref_tmp;
  }
  pointer operator->() const {
    return std::addressof(operator*());
  }
  stashing_reverse_iterator& operator++() {
    --current;
    return *this;
  }
 stashing_reverse_iterator operator++(int) {
    stashing_reverse_iterator tmp = *this;
    --current;
    return tmp;
  }
  stashing_reverse_iterator& operator--() {
    ++current;
    return *this;
  }
  stashing_reverse_iterator operator--(int) {
    stashing_reverse_iterator tmp = *this;
    ++current;
    return tmp;
  }
  stashing_reverse_iterator  operator+ (difference_type n) const {
    return stashing_reverse_iterator(current - n);
  }
  stashing_reverse_iterator& operator+=(difference_type n) {
    current -= n;
    return *this;
  }
  stashing_reverse_iterator  operator- (difference_type n) const {
    return stashing_reverse_iterator(current + n);
  }
  stashing_reverse_iterator& operator-=(difference_type n) {
    current += n;
    return *this;
  }
  // Differs from reverse_iterator::operator[]:
  // 1. const qualifier removed because this function makes use of operator*
  reference operator[](difference_type n) {
    return *(*this + n);
  }
protected:
  Iterator current;
private:
  Iterator deref_tmp;
};
template <class Iterator1, class Iterator2>
bool operator==(
  const stashing_reverse_iterator<Iterator1>& x,
  const stashing_reverse_iterator<Iterator2>& y)
{ return x.base() == y.base(); }
template <class Iterator1, class Iterator2>
bool operator<(
  const stashing_reverse_iterator<Iterator1>& x,
  const stashing_reverse_iterator<Iterator2>& y)
{ return x.base() > y.base(); }
template <class Iterator1, class Iterator2>
bool operator!=(
  const stashing_reverse_iterator<Iterator1>& x,
  const stashing_reverse_iterator<Iterator2>& y)
{ return !(x.base() == y.base()); }
template <class Iterator1, class Iterator2>
bool operator>(
  const stashing_reverse_iterator<Iterator1>& x,
  const stashing_reverse_iterator<Iterator2>& y)
{ return x.base() < y.base(); }
template <class Iterator1, class Iterator2>
bool operator>=(
  const stashing_reverse_iterator<Iterator1>& x,
  const stashing_reverse_iterator<Iterator2>& y)
{ return x.base() <= y.base(); }
template <class Iterator1, class Iterator2>
bool operator<=(
  const stashing_reverse_iterator<Iterator1>& x,
  const stashing_reverse_iterator<Iterator2>& y)
{ return x.base() >= y.base(); }
template <class Iterator1, class Iterator2>
auto operator-(
  const stashing_reverse_iterator<Iterator1>& x,
  const stashing_reverse_iterator<Iterator2>& y) -> decltype(y.base() - x.base())
{ return y.base() - x.base(); }
template <class Iterator>
stashing_reverse_iterator<Iterator> operator+(
  typename stashing_reverse_iterator<Iterator>::difference_type n,
  const stashing_reverse_iterator<Iterator>& x)
{ return stashing_reverse_iterator<Iterator>(x.base() - n); }
template <class Iterator>
stashing_reverse_iterator<Iterator> make_stashing_reverse_iterator(Iterator i)
{ return stashing_reverse_iterator<Iterator>(i); }

用法与reverse_iterator:相同

// prints 5,4,3,2,1, for a sanely implemented number_iterator
std::copy(
  make_stashing_reverse_iterator(number_iterator(5)),
  make_stashing_reverse_iterator(number_iterator(0)),
  std::ostream_iterator<int>(std::cout, ","));