Vector size and capacity

Vector size is simply the number of elements in the vector:

1. Current vector size is queried by size() member function. Convenience empty() function returns true if size is 0:

   vector<int> v = { 1, 2, 3 }; // size is 3
   const vector<int>::size_type size = v.size();
   cout << size << endl; // prints 3
   cout << boolalpha << v.empty() << endl; // prints false
   
2. Default constructed vector starts with a size of 0:

   vector<int> v; // size is 0
   cout << v.size() << endl; // prints 0
   
3. Adding N elements to vector increases size by N (e.g. by push_back(), insert() or resize() functions).
4. Removing N elements from vector decreases size by N (e.g. by pop_back(), erase() or clear() functions).
5. Vector has an implementation-specific upper limit on its size, but you are likely to run out of RAM before reaching it:

   vector<int> v;
   const vector<int>::size_type max_size = v.max_size();
   cout << max_size << endl; // prints some large number
   v.resize( max_size ); // probably won't work
   v.push_back( 1 ); // definitely won't work
   

   Common mistake: size is not necessarily (or even usually) int:

   vector<int> v_bad( N, 1 ); // constructs large N size vector
   for (int i = 0; i < v_bad.size(); ++i) { // size is not supposed to be int!
       do_something( v_bad[i] );
   }

   Vector capacity differs from size. While size is simply how many elements the vector currently has, capacity is for how many elements it allocated/reserved memory for. That is useful, because too frequent (re)allocations of too large sizes can be expensive.

6. Current vector capacity is queried by capacity() member function. Capacity is always greater or equal to size:

   vector<int> v = { 1, 2, 3 }; // size is 3, capacity is >= 3
   const vector<int>::size_type capacity = v.capacity();
   cout << capacity << endl; // prints number >= 3
   
7. You can manually reserve capacity by reserve( N ) function (it changes vector capacity to N):

   // !!!bad!!!evil!!!
   vector<int> v_bad;
   for( int i = 0; i < 10000; ++i ) {
       v_bad.push_back( i ); // possibly lot of reallocations
   }
   
   // good
   vector<int> v_good;
   v_good.reserve( 10000 ); // good! only one allocation
   for( int i = 0; i < 10000; ++i ) {
       v_good.push_back( i ); // no allocations needed anymore
   }
8. You can request for the excess capacity to be released by shrink_to_fit() (but the implementation doesn’t have to obey you). This is useful to conserve used memory:

   vector<int> v = { 1, 2, 3, 4, 5 }; // size is 5, assume capacity is 6
   v.shrink_to_fit(); // capacity is 5 (or possibly still 6)
   cout << boolalpha << v.capacity() == v.size() << endl; // prints likely true (but possibly false)
   

   Vector partly manages capacity automatically, when you add elements it may decide to grow. Implementers like to use 2 or 1.5 for the grow factor (golden ratio would be the ideal value - but is impractical due to being rational number). On the other hand vector usually do not automatically shrink. For example:

   vector<int> v; // capacity is possibly (but not guaranteed) to be 0
   v.push_back( 1 ); // capacity is some starter value, likely 1
   v.clear(); // size is 0 but capacity is still same as before!
   
   v = { 1, 2, 3, 4 }; // size is 4, and lets assume capacity is 4.
   v.push_back( 5 ); // capacity grows - let's assume it grows to 6 (1.5 factor)
   v.push_back( 6 ); // no change in capacity
   v.push_back( 7 ); // capacity grows - let's assume it grows to 9 (1.5 factor)
   // and so on
   v.pop_back(); v.pop_back(); v.pop_back(); v.pop_back(); // capacity stays the same