Help me understand std::erase

Question

In the book 'C++ In A Nutshell', there is the following example code

std::vector<int> data
...
std::erase(std::remove(data.begin(), data.end(), 42),
  data.end());

I thought that 'erase' was a member function, so shouldn't that be 'data.erase' rather than 'std::erase'? Is there some way the c++ compiler can tell what member you wanted to call a member function on, or did the book omit any documentation of an erase template function, or is the example wrong?

Answer 1

erase is a member function. The sample provided is incorrect.

Answer 2

Edit: Sorry their is no generic erase, just double checked

Answer 3

There is no std::erase. std::map::erase, std::list::erase exists. But no std::erase exists.

Look this question about phantom std::erase.

Answer 4

Yes, erase is a member function, so it should be data.erase() instead of std::erase().

Answer 5

You observation is correct. 'Erase' should be a member function. Only a member function on a container can change the memory size of that container.

Answer 6

There is an algorithm called std::remove. And calling erase on the data structure. remove moves all the elements to be erased to the end of the iterator range and returns the first element to be removed. Returns end() if the element could not be found.

So then you call erase on range starting with the return value of std::remove and the end of the data structure.

See: http://www.sgi.com/tech/stl/remove.html

Note that remove won't work with ordered datastructures, as the elements can't be rearranged.

remove is linear, and so would be removing a vector from up to the end. As it wouldn't need to bubble up the elements after the elements to be removed.

std::vector<int> data
...
data.erase(std::remove(data.begin(), data.end(), 42), data.end())

compared to something like this which is O(N**2)

std::vector<int> data
...
for ( i = data.begin(), i != data.end(); ++i ) {
  if ( *i == 42 ) data.erase( i ) ;
}