In article <478810E0.7030407@providere-consulting.com>, Steven Watanabe <watanabesj@gmail.com> wrote:
AMDG
nicola wrote:
Hi, I have a class modeling a record-based read-only binary file and I have written a random access iterator based on iterator_facade to iterate over the records. The class looks like this:
class my_binary_file {
<snip>
explicit my_binary_file_iterator(my_binary_file* ff, unsigned int n) : ff_ptr_(ff), rec_num_(n) { }
my_binary_file* is *not* const here.
Thanks for pointing that out. In the tutorial example at http://www.boost.org/libs/iterator/doc/iterator_facade.html#tutorial-exam ple sect. "A constant node_iterator", shouldn't the constructor's parameter of const_node_iterator be const? That is, explicit const_node_iterator(const /* <== ? */ node_base* p) : m_node(p) {}
<snip>
};
public: typedef my_binary_file_iterator iterator;
iterator begin() *const* { return iterator(this, 0); }
this is a pointer to const. The constructor of iterator requires a pointer to non-const.
There is more than that, however: as you can see in my original post, get_record() and seekg() are not const and that is a problem because they are invoked by dereference(). I can compile successfully if I const-qualify them, but this is not correct in my opinion. The fact is, they do change the state of the object in a visible way. That's the implementation of get_record() that I had omitted in the original post (but that you could have easily guessed): bool get_record(key_type& k, value_type& v) /* const ??? */ { if (fs_.read(reinterpret_cast<char*>(&k), key_size_)) { if (fs_.read(reinterpret_cast<char*>(&v), sizeof(value_type))) { return true; } } return false; } So, for example, calling get_record() twice results in two consecutive records to be read — not the same record. Moreover, I will be interested, at a later stage, in extending the class so that it can read and *write* files. Is there any better solution than using const member functions for my get_record() and seekg()? Nicola