In Range-v3 one can easily create view of existing containers. For example
#include<range/v3/view/transform.hpp>
#include<cassert>
int main(){
std::vector<double> v = {1,2,3};
auto range = v | ranges::v3::view::transform([](auto x){return x*2;});
assert( range[1] == 4 );
// range[1] = 4; // error cannot write to an l-value
}
These views are naturally read-only. I wonder if there is any way to create read and write view within Range-v3?
I do appreciate that such thing is much more complicated than read-only, and not always possible, but still I wonder if there is a protocol in Range-v3 to use this directly or implement it.
For example, I have seen examples of derived classes from ranges::view_facade
that implement a member function read()
(I cannot find the example again, the documentation of Ranges v3 is really scattered around). What about a write(...)
member function?
I am looking for some hypothetical bi
directional transform
code like this:
#include<range/v3/view/transform.hpp>
#include<cassert>
int main(){
std::vector<double> v = {1,2,3};
auto range = v | ranges::v3::view::bitransform([](double x){return x*2;}, [](double x){return x/2;}); // direct and inverse function for read and write
assert( range[1] == 4 );
range[1] = 8; // error cannot write to an l-value
assert( range[1] == 8 );
assert( v[1] == 4 );
}
You can make transform_view
have elements of proxy type:
#include<vector>
#include<range/v3/view/transform.hpp>
#include<cassert>
struct proxy {
double& x;
operator double() { return x*2; }
void operator=(double y) { x = y / 2; }
};
int main(){
std::vector<double> v = {1,2,3};
auto range = v | ranges::v3::view::transform([](auto& x){return proxy{x};});
assert( range[1] == 4 );
range[1] = 8;
assert( range[1] == 8 );
assert( v[1] == 4 );
}