背景：

最近处理的单个文件，大概有13GB，数量条数约5000万。一次性读人到内存需要选择合适的数据结构对其进行存储。本文对比boost::unordered_map 和 std::map
这两种数据结构在该使用情景下的效率。

代码：

#include "boost/unordered_map.hpp"
#include <iostream>
#include <map>
#include "time.h"

using namespace std;
int main()
{
	{
		time_t first_time = time(0);
		boost::unordered_map<int, int> test_hash;
		for (int i = 0; i < 50000000; i++)
		{
			test_hash.insert(std::pair<int, int>(i, i));
		}
		cout << test_hash.size() << endl;

		time_t second_time = time(0);

		cout << "The insert operate cost: " << second_time - first_time << endl;
		for (int i = 0; i< 50000001; ++i)
		{
			boost::unordered_map<int, int>::iterator iter = test_hash.find(i);
			if (iter == test_hash.end())
			{
				cout << "unordered map find the end!" << endl;
			}
		}
		time_t third_time = time(0);
		cout << "The find operate cost: " << third_time - second_time << endl;
	}

	{
		time_t first_time = time(0);
		std::map<int, int> test_hash;
		for (int i = 0; i < 50000000; i++)
		{
			test_hash.insert(std::pair<int, int>(i, i));
		}
		cout << test_hash.size() << endl;

		time_t second_time = time(0);

		cout << "The insert operate cost: " << second_time - first_time << endl;
		for (int i = 0; i< 50000001; ++i)
		{
			std::map<int, int>::iterator iter = test_hash.find(i);
			if (iter == test_hash.end())
			{
				cout << "map find the end!" << endl;
			}
		}
		time_t third_time = time(0);
		cout << "The find operate cost: " << third_time - second_time << endl;
	}
	return 0;
}

运行结果如下所示：

50000000
The insert operate cost: 34
unordered map find the end!
The find operate cost: 15
50000000
The insert operate cost: 49
map find the end!
The find operate cost: 23

效率上：

boost::unordered_map （34s）插入比map(49s)快。

boost::unordered_map （15s）查找操作比map(23s)快。

内存空间占用上：

boost::unordered_map 内存占用26%。7.6GB*0.26=1.976GB。

map内存占用29%。7.6GB*0.29=2.2GB。

所以，在效率和内存占用上面，boost::unordered_map 都更具有优势。