data() for iterator_range

[olafvdspek@…]

Could you add the data() member to iterator_range?

[Steven Watanabe]

What exactly is iterator_range::data supposed to return?

[Olaf van der Spek <olafvdspek@…>]

return empty() ? NULL : &front();

[Steven Watanabe]

This is rather dangerous. It only works if the iterator points to contiguous storage.

[Olaf van der Spek <olafvdspek@…>]

Right, it should only be provided in that case.

[rhys.ulerich@…]

How can a member method be provided only under certain runtime conditions? It seems the existing class methods do what one needs and that data() could be written as a free function.

[Olaf van der Spek <olafvdspek@…>]

Replying to rhys.ulerich@…:

How can a member method be provided only under certain runtime conditions?

It's not a runtime condition.

It seems the existing class methods do what one needs and that data() could be written as a free function.

A free function that'd work for all containers / ranges that support random access would be fine too.

[Olaf van der Spek <olafvdspek@…>]

Except that C++11 has data() for array, string, vector and probably others, so there's probably a good reason against a free function.

[rhys.ulerich@…]

Referring to...

return empty() ? NULL : &front();

This is rather dangerous. It only works if the iterator points to contiguous storage.

Right, it should only be provided in that case.

I asked...

How can a member method be provided only under certain runtime conditions?

It's not a runtime condition.

An iterator_range knowing that it represents a contiguous region of memory seems like a runtime condition to me. What am I missing?

[Olaf van der Spek <olafvdspek@…>]

Replying to rhys.ulerich@…:

An iterator_range knowing that it represents a contiguous region of memory seems like a runtime condition to me. What am I missing?

It's a compile time condition. True for pointers, arrays, array, vector and string. False for a lot of other things.

[rhys.ulerich@…]

An iterator_range knowing that it represents a contiguous region of memory seems like a runtime condition to me. What am I missing?

It's a compile time condition. True for pointers, arrays, array, vector and string. False for a lot of other things.

Ahh.. Good point. My mistake.

[viboes]

Replying to Olaf van der Spek <olafvdspek@…>:

Right, it should only be provided in that case.

How the user could know that the function is provided only under some special circumstances? Could you show a real use case?

[anonymous]

I'm sure Google can give you enough real use cases for data(). It's a standard function.

[Neil Groves]

While a member data() function exists for some containers in the standard library it does not make sense to conditionally add this to iterator_range.

If you know enough about the iterator that this function can work then it is trivial to implement a non-member function that is more broadly applicable.

[anonymous]

But ISO C++ has data() as a member function, not as a non-member function.