I would always prefer try_emplace over emplace. A Crucial difference is that try_emplace will not construct the object associated with the key,if the key already exists.This will boost the performance in case objects of that type are expensive to create
For example the below code (Example from https://github.com/PacktPublishing/Cpp17-STL-Cookbook/blob/master/Chapter02/efficient_insert_or_reassign_to_map.cpp)
#include <iostream>
#include <functional>
#include <list>
#include <map>
using namespace std;
struct billionaire {
string name;
double dollars;
string country;
};
int main()
{
list<billionaire> billionaires {
{"Bill Gates", 86.0, "USA"},
{"Warren Buffet", 75.6, "USA"},
{"Jeff Bezos", 72.8, "USA"},
{"Amancio Ortega", 71.3, "Spain"},
{"Mark Zuckerberg", 56.0, "USA"},
{"Carlos Slim", 54.5, "Mexico"},
// ...
{"Bernard Arnault", 41.5, "France"},
// ...
{"Liliane Bettencourt", 39.5, "France"},
// ...
{"Wang Jianlin", 31.3, "China"},
{"Li Ka-shing", 31.2, "Hong Kong"}
// ...
};
map<string, pair<const billionaire, size_t>> m;
for (const auto &b : billionaires) {
auto [iterator, success] = m.try_emplace(b.country, b, 1);
if (!success) {
iterator->second.second += 1;
}
}
for (const auto & [key, value] : m) {
const auto &[b, count] = value;
cout << b.country << " : " << count << " billionaires. Richest is "
<< b.name << " with " << b.dollars << " B$\n";
}
}
For the above code
m.try_emplace(b.country, b, 1);
if there unsuccessful insertion the pairs will not get constructed which adds to the performance