Igorjan94's blog

By Igorjan94, history, 8 months ago, In English,

C++17 is now available on codeforces, community wants new edition of C++ tricks by Swift, so, let's start!
Disclaimer: I have done only few examples of new features, which in my opinion are related to competitive programming. Feel free to comment and provide more real-world examples or ask to elaborate some features with more examples or explanations.

Fold expressions

  • I think that everybody knows, what reduce or fold means, but a c++11 example:
vector<int> v = {1, 3, 5, 7};
int res = accumulate(v.begin(), v.end(), 0, [](int a, int b) { return a + b; });
cout << res; // 16
  • In C++17 there is also folding support for a template parameters list. It has the following syntax:
(pack op ...)
(... op pack)
(pack op ... op init)
(init op ... op pack)
  • For example, implement a template function that takes a variable number of parameters and calculates their sum. Before C++17 we cannot do this without explicit first argument:
//C++14
auto Sum()
{
    return 0;
}

template<typename Arg, typename... Args>
auto Sum(Arg first, Args... rest)
{
    return first + Sum(rest...);
}

cout << Sum(1, 2, 3, 4, 5); // 15
//C++17
template<typename... Args>
auto Func(Args... args)
{
    return (args + ...);
}

cout << Func(1, 2, 3, 4, 5); // 15
  • This is useful, when we use comma as op:
// C++17
template<typename T, typename... Args>
void pushToVector(vector<T>& v, Args&&... args)
{
    (v.push_back(forward<Args>(args)), ...);
    //This code is expanded into a sequence of expressions separated by commas as follows:
    //  v.push_back(forward<Args_1>(arg1)),
    //  v.push_back(forward<Args_2>(arg2)),
    //  ....
}

vector<int> v;
pushToVector(v, 1, 4, 5, 8);
  • And my favourite example:
//C++17
template<typename... Args>
void readln(Args&... args)
{
    ((cin >> args), ...);
}

template<typename... Args>
void writeln(Args... args)
{
    ((cout << args << " "), ...);
}

int x;
double y;
readln(x, y); // enter 100 500.1234
writeln(x, "some string", y); // 100 some string 500.1234
  • Note: brackets are meaningfull

Class template argument deduction

template<typename T>
struct point
{
    T x;
    T y;
    point(T x, T y) : x(x), y(y) {}
};

//C++11
pair<int, double> p1 = {14, 17.0}
point<int> u = {1, 2};

//C++17
pair p2 = {14, 17.0}
point v = {1, 2};

If struct is complex, there is a possibility to write deduction guides ourselves, for instance:

template<typename T, typename U>
struct S
{
    T first;
    U second;
};

// My deduction guide
template<typename T, typename U>
S(const T &first, const U &second) -> S<T, U>;

Note: the compiler is able to create deduction guide automatically from a constructor, but in this example, the structure S has no constructor, so, we define deduction guide manually.

*this capture in lambda expressions

I don't think this is useful in CP, but who knows:

struct someClass
{
    int x = 0;

    void f() const
    {
        cout << x << '\n';
    }

    void g()
    {
        x++;
    }

    // C++14
    void func()
    {
        auto lambda1 = [self = *this]() { self.f(); };
        auto lambda2 = [self = *this]() mutable { self.g(); };
        lambda1();
        lambda2();
    }

    // C++17
    void funcNew()
    {
        auto lambda1 = [*this]() { f(); };
        auto lambda2 = [*this]() mutable { g(); };
        lambda1();
        lambda2();
    }
};

Article about mutable keyword.

Structured bindings

  • The most useful syntax sugar for decomposition of objects.
template<typename T>
struct point
{
    T x;
    T y;
    point(T x, T y) : x(x), y(y) {}
};

vector<point<int>> points = {{0, 0}, {1, 0}, {1, 1}, {1, 0}};
//C++11
for (auto& point : points)
{
    int x, y;
    tie(x, y) = point;
    //...Some compex logic with x and y
}

//C++17
for (auto& [x, y] : points)
{
    //...Some compex logic with x and y
}
  • Iterating over map:
map<int, string> m;
for (auto [key, value] : m)
    cout << "key: " << key << '\n' << "value: " << value << '\n';
  • A good example of usage is problem 938D - Buy a Ticket. Code with structured bindings (Dijkstra algo) is much more readable and understandable: compare 35474147 and 35346635.
while (!q.empty())
{
    auto [dist, u] = *q.begin();
    q.erase(q.begin());
    used[u] = true;
    for (auto& [w, v] : g[u])
        if (!used[v] && d[v] > dist + 2 * w)
            q.erase({d[v], v}),
            d[v] = dist + 2 * w,
            q.insert({d[v], v});
}

Initializer in if and switch

set<int> s;

if (auto [iter, ok] = s.insert(42); ok)
{
    //...
}
else
{
    //`ok` and `iter` are available here
}
//But not here

New attributes

  • [[fallthrough]] attribute indicates that the break operator inside a case block is missing intentionally:
int requests, type;
cin >> requests;
for (int q = 0; q < requests; ++q)
    switch (cin >> type; type) //Initializer in switch
    {
        case 1:
            int l, r;
            cin >> l >> r;
            //proceed request of first type
            break;
        case 2:
            [[fallthrough]];
            //Compiler warning will be supressed
        case 3:
            int value;
            cin >> value;
            //Proceed requests of second and third types.
    }
  • [[nodiscard]] attribute is used to indicate that the return value of the function should not be ignored and can be also applied to data types.

std::optional

optional<int> findPath(graph g, int from, int to)
{
    //Find path from `from` to `to`
    if (d[to] != INF)
        return d[to];
    return {}
}

//We can check if value exists
if (auto dist = findPath(...); dist.hasValue())
    cout << dist.value(); //And get it
else
    cout << -1;

//Or use defaultValue if value is not set
cout << findPath(...).value_or(-1); //Prints distance if path exists and -1 otherwise

Non-constant string::data

For C-lovers:

string str = "hello";
char *p = str.data();
p[0] = 'H';
cout << str; // Hello

Free functions std::size, std::data and std::empty

In addition to the already existing free functions std::begin, std::end and others, some new free functions appeared, such as: std::size, std::data and std::empty:

vector<int> v = { 3, 2, 5, 1, 7, 6 };

size_t sz = size(v);
bool empty = empty(v);
auto ptr = data(v);

std::clamp

Returns x if it is in the interval [low, high] or, otherwise, the nearest value:

cout << clamp(7, 0,  10); //7
cout << clamp(7, 0,  5);  //5
cout << clamp(7, 10, 50); //10

I think that it is convenient function, but it'll be difficult to call it in mind during contest :)

GCD and LCM!

cout << gcd(24, 60); // 12
cout << lcm(8, 10);  // 40

The return value from emplace_back

vector<int> v = { 1, 2, 3 };

auto &r = v.emplace_back(10);
r = 42;
//v now contains {1, 2, 3, 42}

std::map functions:

  • Extract (and even change key!!!)
map<int, string> myMap{ { 1, "Gennady" }, { 2, "Petr" }, { 3, "Makoto" } };
auto node = myMap.extract(2);
node.key() = 42;
myMap.insert(move(node));

// myMap: {{1, "Gennady"}, {42, "Petr"}, {3, "Makoto"}};

Note: Extract is the only way to change a key of a map element without reallocation

Complexity:
extract(key): doc
extract(iterator): O(1) amortized doc

  • Merge
map<int, string> m1{ { 1, "aa" }, { 2, "bb" }, { 3, "cc" } }; 
map<int, string> m2{ { 4, "dd" }, { 5, "ee" }, { 6, "ff" } };
m1.merge(m2);
// m1: { {1, "aa"}, {2, "bb"}, {3, "cc"}, {4, "dd"}, {5, "ee"}, {6, "ff"} }
// m2: {}

Compexity: doc

  • To figure out if the insert or update occurred, we had to first look for the element, and then apply the operator[]. Now we had insert_or_assign:
map<int, string> m;
m.emplace(1, "aaa");
m.emplace(2, "bbb");
m.emplace(3, "ccc");

auto [it1, inserted1] = m.insert_or_assign(3, "ddd");
cout << inserted1; // 0

auto [it2, inserted2] = m.insert_or_assign(4, "eee");
cout << inserted2; // 1

Complexity: doc

More rigorous evaluation order of expressions

And in general c++17 introduces new rules, defining more strictly the evaluation order of expressions:

  • Postfix expressions are evaluated from left to right (including function calls and access to objects members)
  • Assignment expressions are evaluated from right to left.
  • Operands of operators << and >> are evaluated from left to right.

Thus, as it is mentioned in the proposal for the standard, in the following expressions a is now guaranteed to be evaluated first, then b, then c:

a.b
a->b
a->*b
a(b1, b2, b3)
b @= a
a[b]
a << b << c
a >> b >> c

Note: the evaluation order between b1, b2, b3 is still not defined.

P.S.: All materials are adopted with my examples from here
P.P.S.: I don't think my english is poor, but please PM me about grammar or other mistakes to make this article better!

 
 
 
 
  • Vote: I like it  
  • +415
  • Vote: I do not like it  

»
8 months ago, # |
  Vote: I like it +59 Vote: I do not like it

Thank goodness for clamp. No more max of min of max of...

»
8 months ago, # |
  Vote: I like it 0 Vote: I do not like it

Structured bindings are truly awesome!

Thanks for this great article anyway.

Also here is the cppreference link to the page that presents compiler support for new C++ features. These include C++11, C++14, C++17, and later accepted revisions to the standard, as well as various technical specifications.

»
8 months ago, # |
Rev. 3   Vote: I like it 0 Vote: I do not like it
//C++17
template<typename... Args>
void readln(Args&... args)
{
    ((cin >> args), ...);
}

template<typename... Args>
void writeln(Args... args)
{
    ((cout << args << " "), ...);
}

int x;
int y;

Alternatively if you don't want to overkill too hard:

copy(istream_iterator<T>(cin), istream_iterator<T>(), back_inserter(input));

copy(output.begin(), output.end(), ostream_iterator<T>(cout, " "));


Were there some issues hindering __gcd? Or is it just compiler dependent so they introduced gcd as the standard instead?

  • »
    »
    8 months ago, # ^ |
      Vote: I like it 0 Vote: I do not like it

    I think this comment may be the answer for your question in the last sentence.

  • »
    »
    8 months ago, # ^ |
      Vote: I like it 0 Vote: I do not like it

    No, unfortunately ostream_iterator must be instantiated with type:

    vector<int> myvector;
    copy(myvector.begin(), myvector.end(), ostream_iterator<int>(cout, " "));
    

    But my implementation works with different types without any type-mentioning:

    int x;
    double d;
    vector<int> v(10); //and overloaded ostream& operator<<(ostream& os, vector<T> a) and operator>>
    readln(x, d, v);
    writeln(x, d, v);
    
    • »
      »
      »
      8 months ago, # ^ |
      Rev. 2   Vote: I like it 0 Vote: I do not like it

      ah, yes.

      Edit: <T> got considered as a tag :(

      • »
        »
        »
        »
        8 months ago, # ^ |
          Vote: I like it 0 Vote: I do not like it

        Yes, do not forget, that code must be in ```codeBlock``` or `codeLine`, otherwise everything in chevrons disapears :)

»
8 months ago, # |
Rev. 2   Vote: I like it -34 Vote: I do not like it

Oh, come on! I don't know about it.

»
8 months ago, # |
  Vote: I like it +37 Vote: I do not like it

“community wants new edition of C++ tricks by Swift,...”

»
8 months ago, # |
Rev. 2   Vote: I like it 0 Vote: I do not like it

What is the complexity of merging two maps?

»
8 months ago, # |
  Vote: I like it -20 Vote: I do not like it

Can you add the complexity of the new functions?

  • »
    »
    8 months ago, # ^ |
      Vote: I like it 0 Vote: I do not like it

    Added complexity of new map functions

    • »
      »
      »
      8 months ago, # ^ |
        Vote: I like it -22 Vote: I do not like it

      What about others :size(),empty(),data(),gcd(),lcm(),clamp()

      I know than I can find the answer by googling but it wouldn't be better if you added them

      • »
        »
        »
        »
        8 months ago, # ^ |
          Vote: I like it 0 Vote: I do not like it

        :size(), empty(), data() are only wrappers, so they have no complexity: complexity is in usual functions .size() and so on
        clamp is literally two comparisons
        gcd is literally __gcd and usual euclidean algo
        lcm is literally one multiplication and division
        So, nothing strange and breaking in these functions, only shortness and convenience

        • »
          »
          »
          »
          »
          8 months ago, # ^ |
            Vote: I like it -22 Vote: I do not like it

          Thanks for explaining that. We should always check the complexity of the new methods.

          Imagine if clamp was using linar search. It would be a TLE tool. (I know that's using two comparisons as you said).

»
8 months ago, # |
  Vote: I like it 0 Vote: I do not like it
Why long is of 4 bytes in Codeforces(C++17) while it is of 8 bytes in Codechef(C++14) ?
               Codeforces           Codechef
            (C++17)(4 bytes)    (C++14)(8 bytes)
LONG_MIN :   -2147483648       -9223372036854775808
LONG_MAX :    2147483647        9223372036854775807
ULONG_MAX:    4294967295       18446744073709551615
»
8 months ago, # |
  Vote: I like it 0 Vote: I do not like it

A typical benefit of deduction guide is that we can directly write tuple(x, y, z) instead of make_tuple(x, y, z), and pair(x, y) instead of make_pair(x, y). Five less characters and the new version looks more compact ;)

  • »
    »
    8 months ago, # ^ |
    Rev. 2   Vote: I like it +16 Vote: I do not like it

    But we can just write {x,y,z} instead of make_tuple(x, y, z) and {x,y} instead of make_pair(x, y) and that's 10 and 9 characters less respectively.

    • »
      »
      »
      8 months ago, # ^ |
        Vote: I like it +15 Vote: I do not like it

      not in every context

      • »
        »
        »
        »
        8 months ago, # ^ |
          Vote: I like it 0 Vote: I do not like it

        Would You mind giving an example of such a context?It will be helpful.Thank you.

        • »
          »
          »
          »
          »
          8 months ago, # ^ |
            Vote: I like it +5 Vote: I do not like it

          Case 1:

          auto p = {1, 2};
          cout << p.first << " " << p.second << "\n"; // c++ treats p as an initializer_list<int> and fails in finding its "first" field
          

          Case 2:

          auto p = pair{1, 2};
          cout << p.first << " " << p.second << "\n"; // stdout: 1 2
          

          Case 3:

          auto p = pair(1, 2);
          cout << p.first << " " << p.second << "\n"; // stdout: 1 2
          
          • »
            »
            »
            »
            »
            »
            8 months ago, # ^ |
            Rev. 4   Vote: I like it 0 Vote: I do not like it

            Thanks a lot for pointing out
            When i was writing above comment what i had in my mind was

            pair p={1,2};//Only in c++17 by template argument deduction
            cout << p.first <<" "<<p.second<<"\n"; // stdout: 1 2
            


        • »
          »
          »
          »
          »
          8 months ago, # ^ |
          Rev. 2   Vote: I like it 0 Vote: I do not like it

          Comparisons

          if (p == {x, y}) ... 
          

          doesn't work

»
8 months ago, # |
  Vote: I like it 0 Vote: I do not like it

Auto comment: topic has been updated by Igorjan94 (previous revision, new revision, compare).

»
6 months ago, # |
Rev. 2   Vote: I like it 0 Vote: I do not like it

Here are two more interesting things that should be noted for completeness:

  1. gcd does not show the same behaviour as __gcd, gcd of negative numbers is returned as the positive counterpart. Similarly for lcm, LCM of negative numbers is a positive value. So gcd(-2, -10) = 2, lcm(-2, -10) = 10. Verifiable here: 36998215

  2. random_shuffle no longer supports custom generators(for example, using mt19937 etc.), shuffle, however is supported. More about this.