I'm trying to catch up on C++11 and all the great new features. I'm a bit stuck on lambdas.
Here's the code I was able to get to work:
#include <iostream>
#include <cstdlib>
#include <vector>
#include <string>
#include <functional>
using namespace std;
template<typename BaseT, typename Func>
vector<BaseT> findMatches(vector<BaseT> search, Func func)
{
vector<BaseT> tmp;
for(auto item : search)
{
if( func(item) )
{
tmp.push_back(item);
}
}
return tmp;
}
void Lambdas()
{
vector<int> testv = { 1, 2, 3, 4, 5, 6, 7 };
auto result = findMatches(testv, [] (const int &x) { return x % 2 == 0; });
for(auto i : result)
{
cout << i << endl;
}
}
int main(int argc, char* argv[])
{
Lambdas();
return EXIT_SUCCESS;
}
What I would like to have is this:
template<typename BaseT>
vector<BaseT> findMatches(vector<BaseT> search, function <bool (const BaseT &)> func)
{
vector<BaseT> tmp;
for(auto item : search)
{
if( func(item) )
{
tmp.push_back(item);
}
}
return tmp;
}
Basically I want to narrow down the possible lambdas to a sensible subset of functions.
What am I missing? Is this even possible? I'm using GCC/G++ 4.6.
Stephan T. Lavavej explains why this doesn't work in this video. Basically, the problem is that the compiler tries to deduce BaseT from both the std::vector and the std::function parameter. A lambda in C++ is not of type std::function, it's an unnamed, unique non-union type that is convertible to a function pointer if it doesn't have a capture list (empty []). On the other hand, a std::function object can be created from any possible type of callable entity (function pointers, member function pointers, function objects).
Note that I personally don't understand why you would want to limit the incoming functors to that specific signature (in addition to the fact that indirection through a polymorphic function wrapper, like std::function, is by far more inefficient than a direct call to a functor (which may even be inlined)), but here's a working version. Basically, it disables argument deduction on the std::function part, and only deduces BaseT from the std::vector argument:
template<class T>
struct Identity{
typedef T type;
};
template<typename BaseT>
vector<BaseT> findMatches(vector<BaseT> search,
typename Identity<function<bool (const BaseT &)>>::type func)
{
vector<BaseT> tmp;
for(auto item : search)
{
if( func(item) )
{
tmp.push_back(item);
}
}
return tmp;
}
Live example on Ideone.
Another possible way would be to not restrict the functor type directly, but indirectly through SFINAE:
template<class T, class F>
auto f(std::vector<T> v, F fun)
-> decltype(bool(fun(v[0])), void())
{
// ...
}
Live example on Ideone.
This function will be removed from the overload set if fun doesn't take an argument of type T& or if the return type is not convertible to bool. The , void() makes f's return type void.
As has been revealed by other posters, this is a template argument deduction for std::function.
One intuitive way to make the second code snippet work is to add your base type when calling the template function: findMatches<int>.
Another way not mentioned by Xeo is using std::is_convertible:
template<typename BaseT, typename FUNC>
vector<BaseT> findMatches(vector<BaseT> search, function <bool (const BaseT &)> func)
{
static_assert(std::is_convertible<FUNC, function<bool (const BaseT &)> >::value, "func must be convertible to ...");
vector<BaseT> tmp;
for(auto item : search)
{
if( func(item) )
{
tmp.push_back(item);
}
}
return tmp;
}
It avoids wrapping lamda into std::function, and provides a cleaner error message.
Related
I would like to access a member of std::vector<std::variant> by index. Considering the following snippet:
struct Data {
using data_types = std::variant<std::basic_string<char>, double, int>;
public:
template <class T>
void push_back(const T& t) {
m_data.push_back(t);
}
private:
std::vector<data_types> m_data;
};
int main()
{
Data d;
d.push_back(0);
d.push_back("string");
d.push_back(3.55);
}
I would like to access the values like d[0] (should return int) or d[1] (should return std::string).
What I have tried so far but what isn't working is to add the following public method to the existing struct:
template <class T>
T& operator[](const size_t &index) {
return std::visit([](const T& value) {
return static_cast<T>(value);
}, m_data[index]);
}
Any ideas how to achieve the desired result?
The type of an expression in C++ cannot depend on runtime parameters; basically it can only depend on types of the arguments, plus non-type template arguments.
So d[0] and d[1] must have the same type, as the type of the pieces of the expression are identical, and there are no non-type template arguments.
std::get<int>(d[0]) vs std::get<double>(d[1]) can differ in type.
std::get<1>(d[0]) vs std::get<2>(d[1]) can differ in type.
std::visit is a mechanism used to get around this; here, we create every a function object call, one for each possible type, and then pick one at runtime to actually call. However, the type returned from the visit still follows the above rule: it doesn't depend on what type is stored in the variant, and every possible type in the variant must have a valid instantiation of the function.
C++ type system is not a runtime type system. It is compile-time. Stuff like variant and dynamic_cast and any give some runtime exposure to it, but it is intentionally minimal.
If you are wanting to print the contents of a variant, you can do this:
std::visit([](auto& x){
std::cout << x;
}, d[0]);
the trick here is that each of the various types of variant have a lambda function body written for them (so they all must be valid). Then, at run time, the one actually in the variant is run.
You can also test the variant and ask if it has a specific type, either via std::get or manually.
bool has_int = std::visit([](auto& x){
return std::is_same_v<int, std::decay_t<decltype(x)>>::value;
}, d[0]);
this gives you a bool saying if d[0] has an int in it or not.
The next bit is getting insane. Please don't read this unless you fully understand how to use variants and want to know more:
You can even extract out the type index of the variant and pass that around as a run time value:
template<auto I>
using konstant_t = std::integral_constant<decltype(I),I>;
template<auto I>
constexpr konstant_t<I> konstant_v{};
template<auto...Is>
using venum_t = std::variant< konstant_t<Is>... >;
template<class Is>
struct make_venum_helper;
template<class Is>
using make_venum_helper_t = typename make_venum_helper<Is>::type;
template<std::size_t...Is>
struct make_venum_helper<std::index_sequence<Is...>>{
using type=venum_t<Is...>;
};
template<std::size_t N>
using make_venum_t = typename make_venum_helper<std::make_index_sequence<N>>::type;
template<std::size_t...Is>
constexpr auto venum_v( std::index_sequence<Is...>, std::size_t I ) {
using venum = make_venum_t<sizeof...(Is)>;
constexpr venum arr[]={
venum( konstant_v<Is> )...
};
return arr[I];
}
template<std::size_t N>
constexpr auto venum_v( std::size_t I ) {
return venum_v( std::make_index_sequence<N>{}, I );
}
template<class...Ts>
constexpr auto venum_v( std::variant<Ts...> const& v ) {
return venum_v< sizeof...(Ts) >( v.index() );
}
now you can do this:
using venum = make_venum_t<3>;
venum idx = venum_v(d[0]);
and idx holds the index of the engaged type in d[0]. This is only somewhat useful, as you still need std::visit to use it usefully:
std::visit([&](auto I) {
std::cout << std::get<I>( d[0] );
}, idx );
(within the lambda, I is a std::integral_constant, which can be constexpr converted to an integer.)
but lets you do some interesting things with it.
To extract a value from variant, use std::get:
struct Data
{
...
template <class T>
T& operator[](size_t index)
{
return std::get<T>(m_data[index]);
}
};
However, because this overloaded operator is a template, you can't use simple operator syntax to call it. Use the verbose syntax:
int main()
{
Data d;
d.push_back(0);
d.push_back("string");
d.push_back(3.55);
std::cout << d.operator[]<double>(2);
}
Or rename it to use a plain name instead of the fancy operator[].
Visitor pattern:
#include <iostream>
#include <string>
#include <variant>
#include <vector>
template <class ...Ts>
struct MultiVector : std::vector<std::variant<Ts...>> {
template <class Visitor>
void visit(std::size_t i, Visitor&& v) {
std::visit(v, (*this)[i]);
}
};
int main() {
MultiVector<std::string, int, double> vec;
vec.push_back(0);
vec.push_back("string");
vec.push_back(3.55);
vec.visit(2, [](auto& e) { std::cout << e << '\n'; });
}
I want to make for_each function for 2D-vector but there is an error:
error: no matching function for call to ‘each(std::vector<std::vector<int> >&, main()::<lambda(int&)>)’
How can I fix it?
#include <iostream>
#include <vector>
#include <functional>
using namespace std;
template <class VarType>
void each(vector<vector<VarType>> &ve, function<void (VarType &)> action) {
for(vector<VarType> &v : ve) {
for(VarType &p : v) {
action(p);
}
}
}
int main() {
vector<vector<int>> map(5);
for(vector<int> &ve : map) {
ve.resize(4);
}
each(map, [](int &val) {
val = 1;
});
}
There are a few solutions. I would recommend to just go for a separate template parameter for the function:
template <class VarType, class Function>
void each(vector<vector<VarType>> &ve, Function action) {
// no more worries
}
The problem is that the compiler cannot figure out the VarType in function<void (VarType&)> from the passed lambda. So doing:
function<void(int&)> action = [](int &val) {
val = 1;
};
each(map, action);
would also work, since this way the type (int&) is already known.
P.S. In C++17 you can do std::function action = [](int &val) { val = 1; };.
While I think taking the function as another template parameter as suggested by #DeiDei is the better solution, here is an alternative:
If you want VarType to be deduced from the first function parameter, then you can make the second parameter a non-deduced context:
template <class VarType>
void each(vector<vector<VarType>> &ve,
typename type_identity<function<void (VarType &)>>::type action) {
for(vector<VarType> &v : ve) {
for(VarType &p : v) {
action(p);
}
}
}
This requires C++20 for std::type_identity and #include<type_traits>, but you can implement your own type_identity easily:
template<typename T>
struct type_identity {
using type = T;
};
This works because everything left to the scope resolution operator :: is a non-deduced context, meaning that the template parameters in it will not be deduced from this function parameter. Your original function fails template argument deduction, because VarType cannot be deduced from the second parameter as the second function argument given in the call does not actually have type std::function<...>.
I'm trying to catch up on C++11 and all the great new features. I'm a bit stuck on lambdas.
Here's the code I was able to get to work:
#include <iostream>
#include <cstdlib>
#include <vector>
#include <string>
#include <functional>
using namespace std;
template<typename BaseT, typename Func>
vector<BaseT> findMatches(vector<BaseT> search, Func func)
{
vector<BaseT> tmp;
for(auto item : search)
{
if( func(item) )
{
tmp.push_back(item);
}
}
return tmp;
}
void Lambdas()
{
vector<int> testv = { 1, 2, 3, 4, 5, 6, 7 };
auto result = findMatches(testv, [] (const int &x) { return x % 2 == 0; });
for(auto i : result)
{
cout << i << endl;
}
}
int main(int argc, char* argv[])
{
Lambdas();
return EXIT_SUCCESS;
}
What I would like to have is this:
template<typename BaseT>
vector<BaseT> findMatches(vector<BaseT> search, function <bool (const BaseT &)> func)
{
vector<BaseT> tmp;
for(auto item : search)
{
if( func(item) )
{
tmp.push_back(item);
}
}
return tmp;
}
Basically I want to narrow down the possible lambdas to a sensible subset of functions.
What am I missing? Is this even possible? I'm using GCC/G++ 4.6.
Stephan T. Lavavej explains why this doesn't work in this video. Basically, the problem is that the compiler tries to deduce BaseT from both the std::vector and the std::function parameter. A lambda in C++ is not of type std::function, it's an unnamed, unique non-union type that is convertible to a function pointer if it doesn't have a capture list (empty []). On the other hand, a std::function object can be created from any possible type of callable entity (function pointers, member function pointers, function objects).
Note that I personally don't understand why you would want to limit the incoming functors to that specific signature (in addition to the fact that indirection through a polymorphic function wrapper, like std::function, is by far more inefficient than a direct call to a functor (which may even be inlined)), but here's a working version. Basically, it disables argument deduction on the std::function part, and only deduces BaseT from the std::vector argument:
template<class T>
struct Identity{
typedef T type;
};
template<typename BaseT>
vector<BaseT> findMatches(vector<BaseT> search,
typename Identity<function<bool (const BaseT &)>>::type func)
{
vector<BaseT> tmp;
for(auto item : search)
{
if( func(item) )
{
tmp.push_back(item);
}
}
return tmp;
}
Live example on Ideone.
Another possible way would be to not restrict the functor type directly, but indirectly through SFINAE:
template<class T, class F>
auto f(std::vector<T> v, F fun)
-> decltype(bool(fun(v[0])), void())
{
// ...
}
Live example on Ideone.
This function will be removed from the overload set if fun doesn't take an argument of type T& or if the return type is not convertible to bool. The , void() makes f's return type void.
As has been revealed by other posters, this is a template argument deduction for std::function.
One intuitive way to make the second code snippet work is to add your base type when calling the template function: findMatches<int>.
Another way not mentioned by Xeo is using std::is_convertible:
template<typename BaseT, typename FUNC>
vector<BaseT> findMatches(vector<BaseT> search, function <bool (const BaseT &)> func)
{
static_assert(std::is_convertible<FUNC, function<bool (const BaseT &)> >::value, "func must be convertible to ...");
vector<BaseT> tmp;
for(auto item : search)
{
if( func(item) )
{
tmp.push_back(item);
}
}
return tmp;
}
It avoids wrapping lamda into std::function, and provides a cleaner error message.
I'm learning C++, and I'm trying to implement a binary search function that finds the first element for which a predicate holds. The function's first argument is a vector and the second argument is a function that evaluates the predicate for a given element. The binary search function looks like this:
template <typename T> int binsearch(const std::vector<T> &ts, bool (*predicate)(T)) {
...
}
This works as expected if used like this:
bool gte(int x) {
return x >= 5;
}
int main(int argc, char** argv) {
std::vector<int> a = {1, 2, 3};
binsearch(a, gte);
return 0;
}
But if I use a lambda function as a predicate, I get a compiler error:
search-for-a-range.cpp:20:5: error: no matching function for call to 'binsearch'
binsearch(a, [](int e) -> bool { return e >= 5; });
^~~~~~~~~
search-for-a-range.cpp:6:27: note: candidate template ignored: could not match 'bool (*)(T)' against '(lambda at
search-for-a-range.cpp:20:18)'
template <typename T> int binsearch(const std::vector<T> &ts,
^
1 error generated.
The above error is generated by
binsearch(a, [](int e) -> bool { return e >= 5; });
What's wrong? Why is the compiler not convinced that my lambda has the right type?
Your function binsearch takes a function pointer as argument. A lambda and a function pointer are different types: a lambda may be considered as an instance of a struct implementing operator().
Note that stateless lambdas (lambdas that don't capture any variable) are implicitly convertible to function pointer. Here the implicit conversion doesn't work because of template substitution:
#include <iostream>
template <typename T>
void call_predicate(const T& v, void (*predicate)(T)) {
std::cout << "template" << std::endl;
predicate(v);
}
void call_predicate(const int& v, void (*predicate)(int)) {
std::cout << "overload" << std::endl;
predicate(v);
}
void foo(double v) {
std::cout << v << std::endl;
}
int main() {
// compiles and calls template function
call_predicate(42.0, foo);
// compiles and calls overload with implicit conversion
call_predicate(42, [](int v){std::cout << v << std::endl;});
// doesn't compile because template substitution fails
//call_predicate(42.0, [](double v){std::cout << v << std::endl;});
// compiles and calls template function through explicit instantiation
call_predicate<double>(42.0, [](double v){std::cout << v << std::endl;});
}
You should make your function binsearch more generic, something like:
template <typename T, typename Predicate>
T binsearch(const std::vector<T> &ts, Predicate p) {
// usage
for(auto& t : ts)
{
if(p(t)) return t;
}
// default value if p always returned false
return T{};
}
Take inspiration from standard algorithms library.
The lambda expression with empty capture list could be implicitly converted to a function pointer. But the function pointer predicate is taking T as its parameter, that need to be deduced. Type conversion won't be considered in template type deduction, T can't be deduced; just as the error message said, candidate template (i.e. binsearch) is ignored.
You can use operator+ to achieve this, it'll convert lambda to function pointer, which will be passed to binsearch later and then T will be successfully deduced[1].
binsearch(a, +[](int e) -> bool { return e >= 5; });
// ~
Of course you could use static_cast explicitly:
binsearch(a, static_cast<bool(*)(int)>([](int e) -> bool { return e >= 5; }));
Note if you change predicate's type to be independent of T, i.e. bool (*predicate)(int), passing lambda with empty capture list would work too; the lambda expression will be converted to function pointer implicitly.
Another solution is changing the parameter type from function pointer to std::function, which is more general for functors:
template <typename T> int binsearch(const std::vector<T> &ts, std::function<bool (typename std::vector<T>::value_type)> predicate) {
...
}
then
binsearch(a, [](int e) -> bool { return e >= 5; });
[1] A positive lambda: '+[]{}' - What sorcery is this?
Why is the compiler not convinced that my lambda has the right type?
Template functions being told to deduce their template parameters do no conversion. A lambda is not a function pointer, so there is no way to deduce the T in that argument. As all function arguments independently deduce their template parameters (unless deduction is blocked), this results in an error.
There are a number of fixes you can do.
You can fix the template function.
template <class T>
int binsearch(const std::vector<T> &ts, bool (*predicate)(T))
Replace the function pointer with a Predicate predicate or Predicate&& predicate and leave the body unchanged.
template <class T, class Predicate>
int binsearch(const std::vector<T> &ts, Predicate&& predicate)
Use deduction blocking:
template<class T>struct tag_t{using type=T;};
template<class T>using block_deduction=typename tag_t<T>::type;
template <class T>
int binsearch(const std::vector<T> &ts, block_deduction<bool (*)(T)> predicate)
optionally while replacing the function pointer with a std::function<bool(T)>.
You can fix it at the call site.
You can pass T manually binsearch<T>(vec, [](int x){return x<0;}).
You can decay the lambda to a function pointer with putting a + in front of it +[](int x) ... or a static_cast<bool(*)(int)>( ... ).
The best option is Predicate one. This is also what standard library code does.
We can also go a step further and make your code even more generic:
template <class Range, class Predicate>
auto binsearch(const Range &ts, Predicate&& predicate)
-> typename std::decay< decltype(*std::begin(ts)) >::type
The -> typename std::decay ... trailing return type part can be eliminated in C++14.
A benefit to this is that if the body also uses std::begin and std::end to find begin/end iterators, binsearch now supports deques, flat C-style arrays, std::arrays, std::strings, std::vectors, and even some custom types.
If you have any control over binsearch I suggest you refactor it:
template <typename T, typename Predicate>
int binsearch(std::vector<T> const& vec, Predicate&& pred) {
// This is just to illustrate how to call pred
for (auto const& el : vec) {
if (pred(el)) {
// Do something
}
}
return 0; // Or anything meaningful
}
Another way for you is to perform type-erasure on your functor objects/function pointers/whatever... by embedding them into an std::function<bool(T const&)>. To do so, just rewrite the function above as:
template <typename T>
int binsearch(std::vector<T> const& vec, std::function<bool(T const&)> pred);
But since template argument deduction does not do any conversion, you need to explicitly feed your function like the following:
auto my_predicate = [](int x) { return true; }; // Replace with actual predicate
std::vector<int> my_vector = {1, 2, 3, 4};
binsearch(my_vector, std::function<bool (int const&)>(my_predicate));
However given the description of your function, it seems it does the same job as std::find_if.
std::vector<int> my_vector = {1, 12, 15, 13, 16};
auto it = std::find_if(std::begin(my_vector), std::end(my_vector),
[](int vec_el) { return !vec_el%5; });
// it holds the first element in my_vector that is a multiple of 5.
if (it != std::end(my_vector)) {
std::cout << *it << std::endl; // prints 15 in this case
}
Note that to do a binary search, you need more than just a predicate: you need a predicate that defines an order on your range AND a target value.
A function pointer and a lambda function is not the same thing.
An object t can not be assigned to predicate where:
bool (*predicate)(int)
and
auto t = [](int e) -> bool { return e >= 5; });
Might as well use std::function<bool(int)>. Your signature will look like this:
template <typename T>
int binsearch(const std::vector<T> &ts, std::function<bool(T)> predicate){
// ...
}
Now that's not a function pointer, You'll need to bind your founction pointers if they are necessary, (I assume you're fine with just lambdas)
Is store function pointers with different parameters in a vector of void pointers.
unordered_map<string, vector<void*> > List;
template <typename T>
void Listen(string Name, function<void(T)> Function)
{
List[Name].push_back(&Function);
}
Then I want to call them, assuming that T is the same type for Fire as used for the Listen.
template <typename T>
void Fire(string Name, T Data)
{
auto Functions = List[Name];
for (auto i = Functions.begin(); i != Functions.end(); ++i)
{
(function<void(T)>)i)(Data);
}
}
But I get a compiler error which reads error C2064: term does not evaluate to a function taking 1 arguments in file ...\vc\include\xrefwrap 431 1.
What am I doing wrong?
For one, you're taking the address of a parameter, here:
List[Name].push_back(&Function);
Then you're trying to convert an iterator object to a std::function object here:
(function<void(T)>)i)
What you trying to do can be done, like this, although it's not pretty, to put it mildly:
unordered_map<string, vector<void*> > List;
template <typename T>
void Listen(string Name, function<void(T)> &Function)
{
List[Name].push_back(&Function);
}
template <typename T>
void Fire(string Name, T Data)
{
auto Functions = List[Name];
for (auto i = Functions.begin(); i != Functions.end(); ++i)
{
function<void(T)> *ptr = *i;
(*ptr) (Data);
}
}
It can break in lot of ways, for example you have no control that the function, registered under some name in Listen is called with the correct argument in Fire - consider calling Listen<int> ("foo", f); and then doing Fire<double> ("foo", 3.14);
Another approach - just pass closures for callbacks:
unordered_map<string, std::vector<function<void()> > > List;
void Listen(string Name, function<void()> Function)
{
List[Name].push_back(Function);
}
void Fire(string Name)
{
auto Functions = List[Name];
for (auto i = Functions.begin(); i != Functions.end(); ++i)
(*i) ();
}
#include <functional>
#include <unordered_map>
#include <memory>
#include <string>
#include <vector>
template<typename T> struct BlockDeduction{typedef T type;};
struct BaseCallback {
virtual ~BaseCallback();
template<typename T>
void DoCall( typename BlockDeduction<T>::type&& t ) const;
};
template<typename T>
struct Callback: BaseCallback
{
std::function<void(T)> func;
Callback( std::function<void(T)> const& f ):func(f) {}
};
template<typename T>
void BaseCallback::DoCall( typename BlockDeduction<T>::type&& t ) const {
Assert( dynamic_cast<Callback<T>const*>(this) );
static_cast<Callback<T>const*>(this).func(std::forward(t));
}
typedef std::unique_ptr<BaseCallback> upCallback;
template<typename T>
upCallback make_callback( std::function<void(T)> const& f ) {
return upCallback( new Callback<T>( f ) );
}
struct Listener {
std::unordered_map< std::string, std::vector<upCallback>> List;
template<typename T>
void Listen( std::string Name, std::function<void(T)> f) {
List[Name].push_back( make_callback(f) );
}
template<typename T>
void Fire( std::string Name, typename BlockDeduction<T>::type&& t ) {
auto callbacks = List.find(Name);
if (callbacks == List.end()) return;
for(auto it = callbacks->second.begin(); it != callbacks->second.end(); ++it) {
if (it +1 = callbacks->second.end())
{
(**it).DoCall<T>( std::forward(t) );
} else {
(**it).DoCall<T>( t );
}
}
}
};
... or something like that.
This stores a copy of the std::function in the map, wrapped up generically. Memory is handled via a unique_ptr. I carefully blocked type deduction at points where the type must be exactly what you used when you installed the Listener (automatic type deduction at that point is rather fragile).
In debug, you'll get an assertion failure if you violate the Name<->type mapping.
Some extra work needs to be done for nullary callbacks. Just write a DoCall that casts BaseCallback to Callback<void>, specialize Callback<void> to be a nullary function wrapper, specialize make_callback on nullary function, and write a Fire(string) method for Listener that calls the bare DoCall.
Or create a struct Empty and use lambdas to wrap nullary functions in function<void(Empty)>, which would involve slightly less code, but would be slower at run-time.