the efficiency about them?

[Darwinnpos]

std::vector<std::vector<std::vector>> a; //ervery vector have n size
std::vector> b; //vector have n ^ 3 size
the efficiency about them?

[Darwinnpos]

my code and why?

#pragma warning(disable:4996)
#include <iostream>
#include <vector>
#include <yas/serialize.hpp>
#include <yas/std_types.hpp>

int main()
{
	std::vector<int> vData1;
	std::vector<std::vector<std::vector<int>>> vData2;
	for (int i{ 0 }; i < 1000000; i++)
	{
		vData1.push_back(i);
	}
	for (int i{ 0 }; i < 100; i++)
	{
		std::vector<std::vector<int>> a;
		for (int j{ 0 }; j < 100; j++)
		{
			std::vector<int> b;
			for (int k{ 0 }; k < 100; k++)
			{
				b.push_back(rand());
			}
			//std::cout << " b size" << b.size() << std::endl;
			a.push_back(b);
		}
		//std::cout << " a size" << a.size() << std::endl;
		vData2.push_back(a);
	}
	//std::cout << " vData2 size" << vData2.size() << std::endl;
	// 序列化
	constexpr std::size_t flags =
		yas::mem // IO type
		| yas::binary; // IO format
	std::cout << "==============TEST START===============\n\n";

	for (int i{ 0 }; i < 1; i++)
	{

		auto start = std::chrono::high_resolution_clock::now();

		auto buf = yas::save<flags>(
			YAS_OBJECT("myobject", vData1)
		);
		yas::load<flags>(buf,
			YAS_OBJECT_NVP("myobject"
				, ("a", vData1)
			)
		);
		auto end = std::chrono::high_resolution_clock::now();

		std::cout << "data type std::vector<int> size=1000 000\n";
		std::cout << "yas time cost: " << std::chrono::duration_cast<std::chrono::microseconds>(end - start).count() / 1000.0 << " ms\n";
		std::cout << "yas data size: " << buf.size/1000000.0 << " M byte\n\n";
		// 反序列化
		start = std::chrono::high_resolution_clock::now();
		auto buf2 = yas::save<flags>(
			YAS_OBJECT("myobject", vData2)
		);
		yas::load<flags>(buf2,
			YAS_OBJECT_NVP("myobject"
				, ("a", vData2)
			)
		);
		end = std::chrono::high_resolution_clock::now();
		std::cout << "data type std::vector<std::vector<std::vector<int>>> size=100 * 100 * 100\n";
		std::cout << "time cost: " << std::chrono::duration_cast<std::chrono::microseconds>(end - start).count() / 1000.0 << " ms\n";
		std::cout << "yas data size: " << buf2.size / 1000000.0 << " M byte\n\n";
	}

	std::cout << "==============TEST END===============\n";
	std::cout << "Hello World!\n";
}

[Darwinnpos]

==============TEST START===============

data type std::vector size=1000 000
yas time cost: 6.962 ms
yas data size: 4.00002 M byte

data type std::vector<std::vector<std::vector>> size=100 * 100 * 100
time cost: 50.75 ms
yas data size: 4.08082 M byte

==============TEST END===============
Hello World!

[niXman]

hmm... will look at this...

[niXman]

@Darwinnpos I can't confirm that.

the results for me:

==============TEST START===============

data type std::vector<int> size=1000 000
yas time cost: 1.188 ms
yas data size: 4.00002 M byte

data type std::vector<std::vector<std::vector<int>>> size=100 * 100 * 100
time cost: 2.728 ms
yas data size: 4.08082 M byte

==============TEST END===============
Hello World!

compiled with -std=c++14 -O2

[Darwinnpos]

I mean why the time cost is different？

[niXman]

I mean why the time cost is different？

ah, got it.
will look at that.