C++ iterators and range-based loops are incredibly useful, however defining iterators still requires a large amount of boilerplate code. The goal of this library is to find alternative and useful ways to use and create C++17 iterators without impacting performance or compiler optimizations.
EasyIterator adds well-known generators and iterator combinators from other languages to C++, such as range
, zip
and enumerate
.
using namespace easy_iterator;
std::vector<int> integers(10);
std::vector<std::string> strings(integers.size());
for (auto i: range(integers.size())) {
integers[i] = i*i;
}
for (auto [i, v, s]: zip(range(integers.size()), integers, strings)) {
s = std::to_string(i) + "^2 = " + std::to_string(v);
}
for (auto [i, s]: enumerate(strings)) {
std::cout << "strings[" << i << "] = \"" << s << "\"" << std::endl;
}
Most iterator boilerplate code is defined in an easy_iterator::IteratorPrototype
base class type.
A possible implementation of the range
iterable is below.
using namespace easy_iterator;
template <class T> struct RangeIterator: public IteratorPrototype<T, dereference::ByValue> {
T increment;
RangeIterator(const T &start):
IteratorPrototype<T, dereference::ByValue>(start),
increment(1) {
}
RangeIterator &operator++(){ RangeIterator::value += increment; return *this; }
};
template <class T> auto range(T end) {
return wrap(RangeIterator<T>(begin), RangeIterator<T>(end));
}
Algorithms can be easily wrapped into iterators by defining a class that defines advance()
and value()
member functions. The code below shows how to define an iterator over Fibonacci numbers.
struct Fibonacci {
unsigned current = 0;
unsigned next = 1;
void advance() {
auto tmp = next;
next += current;
current = tmp;
}
unsigned value() {
return current;
}
};
using namespace easy_iterator;
for (auto [i,v]: enumerate(MakeIterable<Fibonacci>())){
std::cout << "Fib_" << i << "\t= " << v << std::endl;
if (i == 10) break;
}
Algorithms that have an end state can also be defined by returning a the state in the advance()
method. If the initial state can also be undefined, the iterator should define a bool init()
method and inherit from easy_iterator::InitializedIterable
. The code below shows an alternative range
implementation.
template <class T> struct RangeIterator: public easy_iterator::InitializedIterable {
T current, max, step;
RangeIterator(T end): current(0), max(end), step(1) { }
bool advance(){ current += step; return current != max; }
bool init(){ return current != max; }
T value(){ return current; }
};
template <class T> auto range(T end) {
return easy_iterator::MakeIterable<RangeIterator<T>>(end);
}
EasyIterator is a single-header library, so you can simply download and copy the header into your project, or use the Cmake script to install it globally.
Using the CPM dependency manager, you can also include EasyIterator simply by adding the following to your projects' CMakeLists.txt
.
CPMAddPackage("gh:thelartians/[email protected]")
target_link_libraries(myProject EasyIterator)
set_target_properties(myProject PROPERTIES CXX_STANDARD 17)
You can run the tests suite included in this repo with the following commands.
cmake -Htest -Bbuild/test
cmake --build build/test
cmake --build build/test --target test
EasyIterator is designed to come with little or no performance impact compared to handwritten code. For example, using for(auto i: range(N))
loops create identical assembly compared to regular for(auto i=0;i<N;++i)
loops (using clang++ -O2
).
The performance of different methods and approaches can be compared with the included benchmark suite.
You can build and run the benchmark with the following commands:
cmake -Hbenchmark -Bbuild/bench -DCMAKE_BUILD_TYPE=Release
cmake --build build/bench -j8
./build/bench/EasyIteratorBenchmark