Implicit conversion of iterator to reverse_iterator

[bb]

Hi,
Please could you clarify why 'implicit conversion' does not take place
while assigning an iterator to reverse_iterator. However, it happens
while initializing/constructing.

e.g.

typedef std::map<int, std::string> MIS;
MIS m1;

// populating m1

MIS::iterator it = m1.lower_bound(4);

// checking the validity of 'it' etc.

MIS::reverse_iterator rit1(it); // implicit conversion takes place
as expected
MIS::reverse_iterator rit2 = it; // does not compile! why there is
no implicit conversion here?
MIS::reverse_iterator rit3 =
static_cast<MIS::reverse_iterator>(it); // compiles & all looks
fine. Is it legal?

Thanks.

 

[Markus Schoder]

Hi,
Please could you clarify why 'implicit conversion' does not take place
while assigning an iterator to reverse_iterator. However, it happens
while initializing/constructing.

e.g.

typedef std::map<int, std::string> MIS;
MIS m1;

// populating m1

MIS::iterator it = m1.lower_bound(4);

// checking the validity of 'it' etc.

MIS::reverse_iterator rit1(it); // implicit conversion takes place
as expected

This calls the constructor reverse_iterator(Iterator x) explicitly -- no
conversion involved.

MIS::reverse_iterator rit2 = it; // does not compile! why there is
no implicit conversion here?

Because the constructor reverse_iterator(Iterator x) is declared
explicit.

MIS::reverse_iterator rit3 =
static_cast<MIS::reverse_iterator>(it); // compiles & all looks
fine. Is it legal?

This is legal and does work.

 

[dasjotre]

Hi,
Please could you clarify why 'implicit conversion' does not take place
while assigning an iterator to reverse_iterator. However, it happens
while initializing/constructing.

e.g.

typedef std::map<int, std::string> MIS;
MIS m1;

// populating m1

MIS::iterator it = m1.lower_bound(4);

// checking the validity of 'it' etc.

MIS::reverse_iterator rit1(it); // implicit conversion takes place
as expected

this is explicit constructor call.

if implicit conversion was allowed
what would you expect compiler to
do with:

for_each(m1.begin(), m1.end(), do_something());
// followed by
for_each(m1.rbegin(), m1.rend(), do_something());

how woudl you expect for_each to make distinction
between reverse_iterator and iterator?

 

[peter koch]

this is explicit constructor call.

if implicit conversion was allowed
what would you expect compiler to
do with:

for_each(m1.begin(), m1.end(), do_something());
// followed by
for_each(m1.rbegin(), m1.rend(), do_something());

how woudl you expect for_each to make distinction
between reverse_iterator and iterator?

I do not see the problem here. An iterator is not a reverse iterator,
and both functions would be "perfect fits". What could be dangerous
would be to mix iterators and reverse_iterators, but in templated code
where no implicit conversions are made, I see no problems.

/Peter