emplace() vs insert() when using std::unordered_map

The emplace() method constructs the key-value pair directly within the map, while insert() requires creating the pair outside the map and then moving or copying it in.

Using emplace() can be more efficient, especially when the key or value type is expensive to copy or move. Here's an example:

1#include <iostream>
2#include <string>
3#include <unordered_map>
4
5struct User {
6  std::string Name;
7  // Other data members...
8
9  User(const std::string& name) : Name{name} {  
10    std::cout << "User constructed\n";          
11  }                                             
12};
13
14int main() {
15  std::unordered_map<std::string, User> users;
16
17  users.emplace("john", "John");  
18
19  auto [it, success] = users.insert(
20    {"alice", User{"Alice"}}
21  );  
22}

1User constructed
2User constructed

When using emplace(), the User object is constructed directly within the map using the provided arguments. With insert(), we first construct a User object outside the map and then insert it, resulting in an extra construction.

However, if you already have an existing key-value pair object, using insert() can be more convenient and readable.

In summary, prefer emplace() when you have the individual key and value components available, especially for complex types. Use insert() when you already have a key-value pair object ready to be inserted.