Iterate through a map of std::variant - c++

I'm experimenting with C++17's std::variant to store data of multiple types in a map. The use-case here is to have a map of controllers of generic types (but bound by std::variant) that I can iterate through and call methods of.
In below example,
#include <iostream>
#include <map>
#include <variant>
class ControlA {
public:
void specificToA() { std::cout << "A" << std::endl; }
};
class ControlB {
public:
void specificToB() { std::cout << "B" << std::endl; }
};
template<typename T>
class ControlItem{
T* control;
public:
ControlItem() = default;
~ControlItem() = default;
void doStuff() {
if constexpr (std::is_same_v<T, ControlA>) {
control->specificToA();
}
if constexpr (std::is_same_v<T, ControlB>) {
control->specificToB();
}
}
};
class MyClass {
public:
void cycleThroughMap();
std::map<std::string, std::variant<ControlItem<ControlA>, ControlItem<ControlB>>> controlMap;
};
The heuristic method for this would be to get the mapped value of each declared type like:
void MyClass::cycleThroughMap() {
for (auto controlItem : controlMap) {
if (auto control = std::get_if<ControlItem<ControlA>>(&controlItem.second)) {
control->doStuff();
} else if (auto control = std::get_if<ControlItem<ControlB>>(&controlItem.second)) {
control->doStuff();
} else
std::cout << "Unknown type!" << std::endl;
}
}
This works but feels like it's not meant to exist.
Can std::variant be used for this? Is it a bad idea from the start, should I use inheritance and voilĂ ?

Can std::variant be used for this?
Yes. Your code is primed for using a variant effectively. The variant holds types with the same implicit interface. It's a perfect opportunity to use std::visit with a generic lambda.
void MyClass::cycleThroughMap() {
for (auto& [ key, control ] : controlMap) {
std::visit([](auto&& c) {
c.doStuff();
}, control);
}
}
I also took the liberty of replacing the pair access with a structured binding. For some added simplicity.

Another way to structure the code - removes the need for get_if. Comments inline:
#include <map>
#include <variant>
#include <iostream>
class ControlA {
public:
void specificToA() { std::cout << "A" << std::endl; }
};
// consistent free-function interface for each operation type allows ADL lookup
void adlDoStuff(ControlA& c)
{
// but with different implementation details
c.specificToA();
}
class ControlB {
public:
void specificToB() { std::cout << "B" << std::endl; }
};
// consistent free-function interface for each operation type allows ADL lookup
void adlDoStuff(ControlB& c)
{
// but with different implementation details
c.specificToB();
}
template<typename T>
class ControlItem{
T* control;
public:
ControlItem() = default;
~ControlItem() = default;
void doStuff() {
// invoke the adl-friendly free functions.
adlDoStuff(*control);
}
};
class MyClass {
public:
void cycleThroughMap();
std::map<std::string, std::variant<ControlItem<ControlA>, ControlItem<ControlB>>> controlMap;
};
void MyClass::cycleThroughMap() {
// use std::visit. Every type of control will have the .doStuff interface
for (auto&& elem : controlMap) {
std::visit([](auto&& control)
{
control.doStuff();
}, elem.second);
}
}

Related

CRTP used with std::any vs virtual functions

I am trying to create a compile time polymorphism design that will not require virtual functions with all their drawbacks. However I am struggling with creating simple, effective and easy to understand container that can simulate the ability to hold derived class in it's base class container. My previous attempts with compile time variadic vectors were working, but the code was huge mess. This solutions seems cleaner to me. I have simple code that implements basic CTRP. However I created a runtime container that is storing std::any objects and then based on the type of the object, I can define the action that is supposed to be taken. I have few questions.
How does the usage of std::any and subsequent any_cast<>() hinder the performance compared to the usage of virtual functions?
Is the usage of std::any valid in this situation?
Is there a better way to implement such container?
Is there a way to force implementation as it is with virtual functions (by using virtual <type> foo() = 0)?
Is it a good idea to create an object that will be a CRTP handler? So I will not have a function for CRTP call, but an object, that can manage those calls?
Thank you.
Here is the base class:
class base {
private:
base() = default;
friend T;
T& implementation = static_cast<T&>(*this);
public:
auto do_stuff() {
return implementation.do_stuff();
}
};
Here is the implementation:
#include <iostream>
class implementation_a : public base<implementation_a> {
public:
auto do_stuff() {
std::cout << 42 << std::endl;
}
};
class implementation_b : public base<implementation_b> {
public:
auto do_stuff() {
return 420;
}
};
Here's the container:
#include <vector>
#include <any>
class crtp_vector {
private:
std::vector<std::any> vec;
public:
auto begin() {
return vec.begin();
}
auto end() {
return vec.end();
}
auto empty() {
return vec.empty();
}
auto size() {
return vec.size();
}
void clear() {
vec.clear();
}
void push_back(const std::any& val) {
vec.push_back(val);
}
auto emplace_back(const std::any& val) {
vec.emplace_back(val);
}
};
Here's the main:
#include "crtp_container.h"
#include <utility>
/* crtp call handler */
template <typename T>
auto crtp_call(T& val) {
return val.do_stuff();
}
int main() {
crtp_vector vec;
implementation_a A;
implementation_b B;
vec.push_back(A);
vec.push_back(B);
for(auto &member : vec) {
if(member.type().name() == typeid(implementation_a).name()) {
crtp_call(std::any_cast<implementation_a&>(member));
}
else if(member.type().name() == typeid(implementation_b).name()) {
std::cout << crtp_call(std::any_cast<implementation_b&>(member)) << std::endl;
}
else {
std::cerr << "no viable type conversion" << std::endl;
}
}
return 0;
}
You make it way too complicated. The code shown doesn't use base in any way; nothing would change if you simply remove it entirely. Even though you keep saying "CRTP", you aren't actually relying on CRTP for anything.
The code doesn't use the ability of std::any to hold any type; it's only used to hold one of a fixed set of types known at compile time. std::variant is better for this.
All told, the example boils down to this:
class implementation_a {
public:
auto do_stuff() {
std::cout << 42 << std::endl;
}
};
class implementation_b {
public:
auto do_stuff() {
std::cout << 420 << std::endl;
return 420;
}
};
int main() {
implementation_a A;
implementation_b B;
std::vector<std::variant<implementation_a, implementation_b>> vec;
vec.push_back(A);
vec.push_back(B);
for(auto &member : vec) {
std::visit([](auto& elem) { elem.do_stuff(); }, member);
}
return 0;
}
Demo

CRTP vs. vector of base class

I have an issue similar to this C++ vector of CRTP shared pointers but my problem formulation adds the fact that the return type of a function i want to use for all inheriting classes is templated.
In detail lets assume this :
template <class Derived>
class Base {
Derived Value() const {
return static_cast<Derived>(this->Value());
};
};
class ChildDouble : public Base<ChildDouble> {
public:
ChildDouble(double r) : _value(r){};
double Value() const {
return _value;
};
private:
double _value;
};
class ChildString : public Base<ChildDouble> {
public:
ChildString(string s) : _value(s){};
string Value() const {
return _value;
};
private:
string _value;
};
Goal would be to use it somewhat similar as in the following main
void main() {
std::vector<Base*> vec;
vec.push_back(new ChildDouble(3.0));
vec.push_back(new ChildString("Thomas"));
unsigned counter = 0;
for (const auto& e : vec) {
std::cout << "Entry " << counter << " : " << e->Value()
<< std::endl;
counter++;
}
}
The compiler is obviously not happy with this because Base requires a template argument.
Any Ideas how this could be solved? AM I using CRTP here although i should not be using it?
Virtual methods (which is what you'd normally need to get the above working without CRTP) won't work here because the interface is different for Value() in each derived type. Virtual inheritance depends on the signature being the same for everyone, except in a few special cases like with covariant return types. It also won't work because virtual methods can't be templated.
But, you can use std::variant to dynamically dispatch your incompatible interfaces, because it is based on templates. First, define a convenient alias for your variant:
using Child = std::variant<ChildDouble, ChildString>;
And then to use, dispatch with std::visit and a generic lambda:
std::vector<Child> vec;
vec.push_back(ChildDouble(3.0));
vec.push_back(ChildString("Thomas"));
unsigned counter = 0;
for (const auto& e : vec) {
std::visit([&counter](auto&& v) {
std::cout << "Entry " << counter << " : " << v.Value()
<< std::endl;
}, e);
counter++;
}
Demo: https://godbolt.org/z/bENWYW
It doesn't work because the compiler doesn't know which type you want to put in the vector and you need to specified it. If you try vector<Base<double>*>vec; it will works but you can't use the vector with other types like Base, because, it is other type.
The solution is to use std::variant or std::any in place of template.
Now you have an object variant/any the declare value in base will make your life easier.
Also I suggest you:
not to use variables starting with underline '_' because this syntax is used by many internal function of compiler.
not to use raw pointer. use smart_ptr like share_ptr then you don't need to worry to destroy it with delete.
Below the code with the changes:
#include <memory>
#include <vector>
#include <string>
#include <variant>
#include <iostream>
using namespace std;
struct Base {
Base(variant<double, string> val) : value(val) {}
void Print() { //just to ilustrate how it works. Better use ostream
if (holds_alternative<double>(this->value))
cout << get<double>(this->value);
else if (holds_alternative<string>(this->value))
cout << get<string>(this->value);
}
protected:
variant<double, string> value;
variant<double, string> BaseValue() const { return this->value; };
};
struct ChildDouble : public Base {
ChildDouble(double r) : Base(r) {};
double Value() const { return get<double>(this->BaseValue()); }
};
struct ChildString : public Base {
ChildString(string s) : Base(s) {};
string Value() const { return get<string>(this->BaseValue()); };
};
int main() { //must return int not void
vector<shared_ptr<Base>>vec;
vec.emplace_back(new ChildDouble(3.0));
vec.emplace_back(new ChildString("Thomas"));
unsigned counter = 0;
for (const auto& e : vec) {
cout << "Entry " << counter << " : "; e->Print(); cout << endl;
++counter;
}
}

How to instantiate input based strategy pattern

The instantiation of strategy pattern is usually completely overlooked by the examples. Let's assume there is an input that defines which class to use. We'd have something along the lines of:
class Strategy {
Strategy(){}
virtual void runAlgorithm() = 0;
};
class A : public Strategy {
A () {}
static bool isA(char input){ return input == 'A'; }
void runAlgorithm { /* Do A algorithm */ }
};
class B : public Strategy {
B () {}
static bool isB(char input){ return input == 'B'; }
void runAlgorithm { /* Do B algorithm */ }
};
// Other algorithms
Strategy* createStrat(char input){
Strategy* instance;
// Define which algorithm to use
if (A::isA(input)) {
instance = A();
} else if (B::isB(input)) {
instance = B();
} ...
// Run algorithm
instance.runAlgorithm();
return instance;
}
As you can see, if we have multiple different algorithms this if/switch can get pretty big. Is there a pattern to make this code easier to humanly parse (i.e., a for loop and a call) without adding an array of pairs? The question could also be expanded to "How to instantiate an input based strategy pattern?"
Do not limit to this code, as it is just an example.
Ok, if you know all your strategies in advance you can use a very simple meta-programming recursion to unroll your if-else chains automatically. Here we go:
#include <string_view>
#include <iostream>
#include <exception>
#include <memory>
struct Strategy {
Strategy(){}
virtual void runAlgorithm() = 0;
};
struct A : public Strategy {
A () {std::cout << "Creating A" << std::endl;}
static constexpr std::string_view id(){
return std::string_view("A");
}
void runAlgorithm() { /* Do A algorithm */ }
};
struct B : public Strategy {
B () {std::cout << "Creating B" << std::endl;}
static constexpr std::string_view id(){
return std::string_view("B");
}
void runAlgorithm() { /* Do B algorithm */ }
};
struct C : public Strategy {
C () {std::cout << "Creating C" << std::endl;}
static constexpr std::string_view id(){
return std::string_view("C");
}
void runAlgorithm() { /* Do C algorithm */ }
};
// the if else chains are constructed by recursion
template <class Head, class... Tail>
struct factory {
static std::unique_ptr<Strategy> call(std::string id) {
if(Head::id() == id) return std::make_unique<Head>();
else return factory<Tail...>::call(id);
}
};
// specialization to end the recursion
// this throws an exception, but you can adapt it to your needs
template <class Head>
struct factory<Head> {
static std::unique_ptr<Strategy> call(std::string id) {
if(Head::id() == id) return std::make_unique<Head>();
else throw std::invalid_argument("The strategy id you selected was not found.");
}
};
// here is your factory which can create instances of A,B,C based only on the runtime id
using my_factory = factory<A,B,C>;
int main() {
auto Astrategy = my_factory::call("A");
auto Bstrategy = my_factory::call("B");
auto Cstrategy = my_factory::call("C");
my_factory::call("D"); // exception thrown
}
Live code here
Edit
Edited to account for smart pointers and error checking as Jarod42 suggested.
Yes there is a solution for that. There are couple ways to do it. Template as in other class or this way:
class StrategyMaker {
public:
void register(std::funcion<bool(char)> predicate,
std::function<std::unique_ptr<Strategy>(char)> factory) {
mFactories.push_back(std::make_pair(std::move(predicate), std::move(factory)));
}
std::unique_ptr<Strategy> create(char ch) {
for (auto pair & : mFactories) {
if (pair.first(ch)) {
return pair.second(ch);
}
}
return {};
}
private:
std::vector<std::pair<std::funcion<bool(char)>,
std::function<std::unique_ptr<Strategy>(char)>>> mFactories;
};
StrategyMaker maker;
maker.register([](char ch){ return input == 'A'; },
[](char ch){ return std::unique_ptr(new A); });
Once I've seen nice article showing how to make it fully automatic. I've found something like this (I'm not 100% sure this is what I read some time ago).

Simulate constructor behaviour for virtual methods

I am currently working on a small private project using C++ i came up with the following structure:
#include <iostream>
class A
{
std::vector<int> vec;
protected:
virtual bool onAdd(int toAdd) {
// should the 'adding' be suppressed?
// do some A specific checks
std::cout << "A::onAdd()" << std::endl;
return false;
}
public:
void add(int i) {
if(!onAdd(i)) {
// actual logic
vec.push_back(i);
}
}
};
class B : public A
{
protected:
bool onAdd(int toAdd) override {
// do some B specific checks
std::cout << "B::onAdd()" << std::endl;
return false;
}
};
In this example onAdd is basically meant to be a callback for add, but in a more polymorphic way.
The actual problem arises when a class C inherits from B and wants to override onAdd too. In this case the implementation in B will get discarded (i.e. not called) when calling C::add. So basically what I would like to achieve is a constructor-like behaviour where I am able to override the same method in different positions in the class hierarchy and all of those getting called.
My question now is: Is there a possibility/design to achieve this? I am sure that it wouldn't be as easy as cascading constructors, though.
Note: Don't focus too much on the add example. The question is about the callback like structure and not if it makes sense with an add.
I would just call my parents onAdd()
bool C::onAdd(int toAdd) {return my_answer && B::onAdd(toAdd);}
This can be a little confusing if you're expecting other developers to inherit from your base class. But for small private hierarchies it works perfectly.
I sometimes include a using statement to make this more explicit
class C : public B
{
using parent=B;
bool onAdd(int toAdd) override {return my_answer && parent::onAdd(toAdd);}
};
struct RunAndDiscard {
template<class Sig, class...Args>
void operator()(Sig*const* start, Sig*const* finish, Args&&...args)const{
if (start==finish) return;
for (auto* i = start; i != (finish-1); ++i) {
(*i)(args...);
}
(*(finish-1))(std::forward<Args>(args)...);
}
};
template<class Sig, class Combine=RunAndDiscard>
struct invokers {
std::vector<Sig*> targets;
template<class...Args>
decltype(auto) operator()(Args&&...args)const {
return Combine{}( targets.data(), targets.data()+targets.size(), std::forward<Args>(args)... );
}
};
struct AndTogetherResultWithShortCircuit {
template<class Sig, class...Args>
bool operator()(Sig*const* start, Sig*const* finish, Args&&...args)const{
if (start==finish) return true;
for (auto* i = start; i != (finish-1); ++i) {
if (!(*i)(args...)) return false;
}
return (*(finish-1))(std::forward<Args>(args)...);
}
};
This creates a per-instance table of things to do onAdd.
Creating a per-class table is harder; you need to chain your table with your parent type's table, which requires per-class boilerplate.
There is no way to get the C++ compiler to write either the per-instance version, or the per-class version, without doing it yourself.
There are C++20 proposals involving reflection and reification, plus the metaclass proposal, which may involve automating writing code like this (on both a per-instance and per-class basis).
Here is a live example of this technique being tested:
struct AndTogetherResultWithShortCircuit {
template<class Sig, class...Args>
bool operator()(Sig*const* start, Sig*const* finish, Args&&...args)const{
if (start==finish) return true;
for (auto* i = start; i != (finish-1); ++i) {
if (!(*i)(args...)) return false;
}
return (*(finish-1))(std::forward<Args>(args)...);
}
};
class A {
std::vector<int> vec;
protected:
invokers<bool(A*, int), AndTogetherResultWithShortCircuit> onAdd;
public:
void add(int i) {
if (!onAdd(this, i)) {
vec.push_back(i);
}
}
};
class B : public A
{
public:
B() {
onAdd.targets.push_back([](A* self, int x)->bool{
// do some B specific checks
std::cout << "B::onAdd(" << x << ")" << std::endl;
return x%2;
});
}
};
class C : public B
{
public:
C() {
onAdd.targets.push_back([](A* self, int x)->bool{
// do some B specific checks
std::cout << "C::onAdd(" << x << ")" << std::endl;
return false;
});
}
};
When you want to write your own OO-system, you can in C++, but C++ doesn't write it for you.
If you want a generic solution perhaps you could use CRTP with variadic templates instead of runtime polymophism.
Taking inspiration from this answer and this answer:
template<class... OnAdders> class A : private OnAdders... {
std::vector<int> vec;
template<class OnAdder>
bool onAdd(int toAdd){
return static_cast<OnAdder*>(this)->onAdd(toAdd);
}
template<typename FirstOnAdder, typename SecondOnAdder, class... RestOnAdders>
bool onAdd(int toAdd){
if (onAdd<FirstOnAdder>(toAdd))
return true;
return onAdd<SecondOnAdder, RestOnAdders...>(toAdd);
}
public:
void add(int i) {
if (onAdd<OnAdders...>(i))
return;
// actual logic
vec.push_back(i);
}
};
class B {
public:
bool onAdd(int toAdd) {
// do some B specific checks
std::cout << "B::onAdd()" << std::endl;
return false;
}
};
Which you could use like:
A<B,C> a;
a.add(42);
Live demo.
The following solution uses std::function to add each callback during each constructor:
#include <iostream>
#include <vector>
#include <functional>
class A
{
std::vector<int> vec;
protected:
bool onAdd(int toAdd)
{
// do some A specific checks
std::cout << "A::onAdd()" << std::endl;
return true;
}
// vector of callback functions. Initialized with A::onAdd() callback as the first entry
std::vector<std::function<bool(int)>> callbacks{{[this](int toAdd){return onAdd(toAdd); }}};
public:
void add(int i)
{
for(auto& callback : callbacks) {
if(!callback(i))
return;
}
// actual logic
vec.push_back(i);
}
};
class B : public A
{
public:
B()
{
callbacks.emplace_back([this](int toAdd){return onAdd(toAdd); });
}
protected:
bool onAdd(int toAdd)
{
// do some B specific checks
std::cout << "B::onAdd()" << std::endl;
return true;
}
};
class C : public B
{
public:
C()
{
callbacks.emplace_back([this](int toAdd){return onAdd(toAdd); });
}
protected:
bool onAdd(int toAdd)
{
// do some C specific checks
std::cout << "C::onAdd()" << std::endl;
// must also call B::onAdd()
return true;
}
};
int main()
{
C c;
c.add(5);
}
Prints:
A::onAdd()
B::onAdd()
C::onAdd()

Calling parametrised method on list items with different template parameters

I'm trying to store and manipulate a list of template class objects with different parameter types; the template class has two parametrised methods, one returning the parameter type and a void one accepting it as input.
More specifically, I have a template class defined as follows:
template<typename T>
class Test
{
public:
virtual T a() = 0;
virtual void b(T t) = 0;
};
And different specifications of it, such as:
class TestInt : public Test<int>
{
public:
int a() {
return 1;
}
void b(int t) {
std::cout << t << std::endl;
}
};
class TestString : public Test<std::string>
{
public:
std::string a() {
return "test";
}
void b(std::string t) {
std::cout << t << std::endl;
}
};
I'd like to be able to store in one single list different objects of both TestInt and TestString type and loop through it calling one method as input for the other, as in:
for (auto it = list.begin(); it != list.end(); ++it)
(*it)->b((*it)->a());
I've looked into boost::any but I'm unable to cast the iterator to the specific class, because I don't know the specific parameter type of each stored object. Maybe this cannot be done in a statically typed language as C++, but I was wondering whether there could be a way around it.
Just for the sake of completeness, I'll add that my overall aim is to develop a "parametrised observer", namely being able to define an observer (as with the Observer Pattern) with different parameters: the Test class is the observer class, while the list of different types of observers that I'm trying to properly define is stored within the subject class, which notifies them all through the two methods a() and b().
The virtuals have actually no meaning here, since for each T the signatures are distinct.
So it seems you have Yet Another version of the eternal "how can we emulate virtual functions templates" or "how to create an interface without virtual functions":
Generating an interface without virtual functions?
How to achieve "virtual template function" in C++
The first one basically contains an idea that you could employ here.
Here's an idea of what I'd do:
Live On Coliru
#include <algorithm>
#include <iostream>
namespace mytypes {
template <typename T>
struct Test {
T a() const;
void b(T t) { std::cout << t << std::endl; }
};
template <> int Test<int>::a() const { return 1; }
template <> std::string Test<std::string>::a() const { return "test"; }
using TestInt = Test<int>;
using TestString = Test<std::string>;
}
#include <boost/variant.hpp>
namespace mytypes {
using Value = boost::variant<int, std::string>;
namespace detail {
struct a_f : boost::static_visitor<Value> {
template <typename T>
Value operator()(Test<T> const& o) const { return o.a(); }
};
struct b_f : boost::static_visitor<> {
template <typename T>
void operator()(Test<T>& o, T const& v) const { o.b(v); }
template <typename T, typename V>
void operator()(Test<T>&, V const&) const {
throw std::runtime_error(std::string("type mismatch: ") + __PRETTY_FUNCTION__);
}
};
}
template <typename O>
Value a(O const& obj) {
return boost::apply_visitor(detail::a_f{}, obj);
}
template <typename O, typename V>
void b(O& obj, V const& v) {
boost::apply_visitor(detail::b_f{}, obj, v);
}
}
#include <vector>
int main()
{
using namespace mytypes;
using AnyTest = boost::variant<TestInt, TestString>;
std::vector<AnyTest> list{TestInt(), TestString(), TestInt(), TestString()};
for (auto it = list.begin(); it != list.end(); ++it)
b(*it, a(*it));
}
This prints
1
test
1
test
Bonus Points
If you insist, you can wrap the AnyTest variant into a proper class and have a() and b(...) member functions on that:
Live On Coliru
int main()
{
using namespace mytypes;
std::vector<AnyTest> list{AnyTest(TestInt()), AnyTest(TestString()), AnyTest(TestInt()), AnyTest(TestString())};
for (auto it = list.begin(); it != list.end(); ++it)
it->b(it->a());
}
Expanding on my comment above, the simplest what I can currently think of to achieve what you are trying to do - at least as I understood it from your example code - is the following:
/* Interface for your container, better not forget the destructor! */
struct Test {
virtual void operate(void) = 0;
virtual ~Test() {}
};
/* Implementation hiding actual type */
template<typename T>
struct TestImpl : public T, public Test {
void operate(void) {
T::b(T::a());
}
};
/* Actual code as template policies */
struct IntTest {
int a(void) {
return 42;
}
void b(int value) {
std::cout << value << std::endl;
}
};
struct StringTest {
std::string a(void) {
return "Life? Don't talk to me about life.";
}
void b(std::string value) {
std::cout << value << std::endl;
}
};
You would then need to create a container for objects of class Test and fill it with objects of the respective TestImpl<IntTest>, TestImpl<StringTest>, and so on. To avoid object slicing you need reference or pointer semantics, that is std::vector<std::unique_ptr<Test> > for example.
for (auto it = list.begin(); it != list.end(); ++it) {
(*it)->operate();
}