使用bitset实现毫秒级查询

前言

因为业务要求api的一次请求响应时间在10ms以内，所以传统的数据库查询操作直接被排除(网络io和磁盘io)。通过调研，最终使用了bitset,目前已经正常运行了很久

---

bitset介绍

看JDK中的解释简直一头雾水，用我自己的理解概括一下

1. bitset的内部实现是long数组
2. set中每一个位的默认值为false（0）
3. bitset长度按需增长
4. bitset非线程安全

---

bitset关键方法分析

    /**
     * Sets the bit at the specified index to {@code true}.
     *
     * @param  bitIndex a bit index
     * @throws IndexOutOfBoundsException if the specified index is negative
     * @since  JDK1.0
     */
    public void set(int bitIndex) {
        if (bitIndex < 0)
            throw new IndexOutOfBoundsException("bitIndex < 0: " + bitIndex);

        int wordIndex = wordIndex(bitIndex);
        expandTo(wordIndex);

        words[wordIndex] |= (1L << bitIndex); // Restores invariants

        checkInvariants();
    }

设置指定“位”为true，bitIndex参数为非负整数。假设我们执行以下代码,观察上面代码中worIndex，words[wordIndex]值的变化

BitSet bs = new BitSet()
bs.set(0);
bs.set(1);
bs.set(2);
bs.set(3);
bs.set(4);

bitIndex	wordIndex	words[wordIndex]	words[wordIndex]二进制表示
0	0	1	0001
1	0	3	0011
2	0	7	0111
3	0	15	1111
4	0	31	0001 1111

通过上表，我们可以很清晰的根据bitIndex和words[wordIndex]二进制值的对应关系，得到一个结论，即：bitset中每一个long可以表示64个非负整数在bitSet中存在与否。例如：0001可以表示整数0在bitset中存在，1111可以表示整数3,2,1,0在bitset中存在。

---

进入正题，实现bitset毫秒级查询

想象一个场景，我们有一张user表,name唯一。

CREATE TABLE `user` (
  `id` int(11) NOT NULL AUTO_INCREMENT,
  `name` varchar(255) NOT NULL,
  `address` varchar(255) NOT NULL comment '地址',
  `gender` varchar(10) NOT NULL comment '性别',
  `age` varchar(10) NOT NULL,
  PRIMARY KEY (`id`)，
  UNIQUE KEY `name` (`name`)
) ENGINE=InnoDB AUTO_INCREMENT=0 DEFAULT CHARSET=utf8;

假设我们要查询“北京市18岁的女生”，那么对应的sql如下：

select * from `user` where address='北京' and age='18' and gender='girl';

如何使用bitset实现同样的查询呢？

1. 将user表数据加载进内存中
2. 为user表建立address,age,gender维度的bitset索引
3. 根据索引查询数据

1.将user表数据加载进内存中

将user表从数据库读取出来存入List

User实体类：

public class User implements Cloneable {
	private String name;
	private String address;
	private String gender;
	private String age;

	@Override
	public String toString() {
		return "User [name=" + name + ", address=" + address + ", gender=" + gender + ", age=" + age + "]";
	}

	@Override
	public User clone() {
		User user = null;
		try {
			user = (User) super.clone();
		} catch (CloneNotSupportedException e) {
			e.printStackTrace();
		}
		return user;
	}
    //省略get set 方法。。。

2.建立索引

创建bitset索引模型类

public class BitSetIndexModel {
	private String type;//索引类型：address，age，gender
	private ConcurrentHashMap<String, Integer> map;//索引类型和bitSet在bsList中下标的映射关系
	private List<String> list;//索引类型的值集合，例如gender：girl，boy
	private List<BitSet> bsList;//bitset集合

	public BitSetIndex() {
		
	}

	public BitSetIndexModel(String type) {
		this.type = type;
		map = new ConcurrentHashMap<String, Integer>();
		list = new ArrayList<String>();
		bsList = new ArrayList<BitSet>();
	}
	
	public String getType() {
		return type;
	}

	public void setType(String type) {
		this.type = type;
	}

	public Map<String, Integer> getMap() {
		return map;
	}

	public void setMap(ConcurrentHashMap<String, Integer> map) {
		this.map = map;
	}

	public List<String> getList() {
		return list;
	}

	public void setList(List<String> list) {
		this.list = list;
	}

	public List<BitSet> getBsList() {
		return bsList;
	}

	public void setBsList(List<BitSet> bsList) {
		this.bsList = bsList;
	}

	/**
	 *
	 * @param str
	 * @param i
	 */
	public void createIndex(String str, int i) {
		BitSet bitset = null;
		//获取‘str’对应的bitset在bsList中的下标
		Integer index = this.getMap().get(str);
		if (index != null) {
			bitset = this.getBsList().get(index);
			if (bitset == null) {
				bitset = new BitSet();
				this.getBsList().add(index, bitset);
			}
			bitset.set(i, true);// 将str对应的位置为true,true可省略
		} else {
			bitset = new BitSet();
			List<String> list = this.getList();
			list.add(str);
			index = list.size() - 1;
			bitset.set(i, true);
			this.getBsList().add(bitset);
			this.getMap().put(str, index);
		}
	}
	
	/**
	 * 从entity里拿出符合条件的bitset
	 *
	 * @param str
	 * @return
	 */
	public BitSet get(String str) {
		BitSet bitset = null;
		str = str.toLowerCase();
		Integer index = this.getMap().get(str);

		if (index != null) {
			bitset = this.getBsList().get(index);
		} else {
			bitset = new BitSet();
		}
		return bitset;
	}

	/**
	 * bitset的与运算
	 *
	 * @param str
	 * @param bitset
	 * @return
	 */
	public BitSet and(String str, BitSet bitset) {
		if (str != null) {
			str = str.toLowerCase();
			if (bitset != null) {
				bitset.and(get(str));
			} else {
				bitset = new BitSet();
				bitset.or(get(str));
			}
		}
		return bitset;
	}

	/**
	 * bitset的或运算
	 *
	 * @param str
	 * @param bitset
	 * @return
	 */
	public BitSet or(String str, BitSet bitset) {
		if (str != null) {
			str = str.toLowerCase();
			if (bitset != null) {
				bitset.or(get(str));
			} else {
				bitset = new BitSet();
				bitset.or(get(str));
			}
		}
		return bitset;
	}

	/**
	 * 获取bitset值为true的 即 把 bitset翻译为list的索引
	 *
	 * @param bitset
	 * @return
	 */
	public static List<Integer> getRealIndexs(BitSet bitset) {
		List<Integer> indexs = new ArrayList<Integer>();
		if (bitset != null) {
			int i = bitset.nextSetBit(0);
			if (i != -1) {
				indexs.add(i);
				for (i = bitset.nextSetBit(i + 1); i >= 0; i = bitset.nextSetBit(i + 1)) {
					int endOfRun = bitset.nextClearBit(i);
					do {
						indexs.add(i);
					} while (++i < endOfRun);
				}
			}
		}

		return indexs;
	}
	
}

为每一个user对象创建address,gender,age维度索引

public class UserIndexStore {

	private static final String ADDRESS = "address";
	private static final String GENDER = "gender";
	private static final String AGE = "age";
	private BitSetIndexModel address;
	private BitSetIndexModel gender;
	private BitSetIndexModel age;
	private ConcurrentHashMap<Integer, User> userMap;//存储所有的user数据
        private ConcurrentHashMap<String, Integer> nameIndexMap;//name和index映射

	public static final UserIndexStore INSTANCE = getInstance();

	private UserIndexStore() {
		init();
	}

	public static UserIndexStore getInstance() {
		return UserIndexStoreHolder.instance;
	}

	private static class UserIndexStoreHolder {
		private static UserIndexStore instance = new UserIndexStore();
	}

	private void init() {
		this.address = new BitSetIndexModel(ADDRESS);
		this.gender = new BitSetIndexModel(GENDER);
		this.age = new BitSetIndexModel(AGE);
		userMap = new ConcurrentHashMap<Integer, User>();
                nameIndexMap =  new ConcurrentHashMap<String, Integer>();
	}

	/**
	 * 构建索引
	 * @param users 
	 */
	public void createIndex(List<User> users) {
		if (users != null && users.size() > 0) {
			for (int index = 0; index < users.size(); index++) {
				User user = users.get(index);
				createIndex(user, index);
			}
		}
	}

	private void createIndex(User user, int index) {
		getAddress().update(user.getAddress(), index);
		getGender().update(user.getGender(), index);
		getAge().update(user.getAge(), index);
		this.userMap.put(index, user);
		this.nameIndexMap.put(user.getName(), index);
	}

	public BitSet query(String address, String gender, String age) {
		BitSet bitset = null;
		bitset = getAddress().and(address, bitset);
		bitset = getGender().and(gender, bitset);
		bitset = getAge().and(age, bitset);
		return bitset;
	}

	public User findUser(Integer index) {
		User user = this.userMap.get(index);
		if (user != null) {
			return user.clone();//可能会对user做修改操作,要保证内存原始数据不变
		}
		return null;
	}

	public BitSetIndexModel getAddress() {
		return address;
	}

	public void setAddress(BitSetIndexModel address) {
		this.address = address;
	}

	public BitSetIndexModel getGender() {
		return gender;
	}

	public void setGender(BitSetIndexModel gender) {
		this.gender = gender;
	}

	public BitSetIndexModel getAge() {
		return age;
	}

	public void setAge(BitSetIndexModel age) {
		this.age = age;
	}

}

3.测试bitset

public class BitSetTest {

	public static void main(String[] args) {
		List<User> users = buildData();
		UserIndexStore.getInstance().createIndex(users);
		ExecutorService executorService = Executors.newFixedThreadPool(20);
		int num = 2000;
		long begin1 = System.currentTimeMillis();
		for (int i = 0; i < num; i++) {
			Runnable syncRunnable = new Runnable() {
				@Override
				public void run() {
					List<Integer> indexs = BitSetIndexModel.getRealIndexs(UserIndexStore.getInstance().query("北京", "girl", "18"));
					for (Integer index : indexs) {
						UserIndexStore.getInstance().findUser(index);
					}
				}
			};
			executorService.execute(syncRunnable);
		}
		executorService.shutdown();
		while (true) {
			try {
				if (executorService.awaitTermination(1, TimeUnit.SECONDS)) {
					System.out.println("单次查询时间为：" + (System.currentTimeMillis() - begin1) / num + "ms");
					break;
				}
			} catch (InterruptedException e) {
				e.printStackTrace();
			}
		}
	}

	private static List<User> buildData() {
		String[] addresss = { "北京", "上海", "深圳" };
		String[] ages = { "16", "17", "18" };
		List<User> users = new ArrayList<>();
		for (int i = 0; i < 200000; i++) {
			User user = new User();
			user.setName("name" + i);
			int rand = ThreadLocalRandom.current().nextInt(3);
			user.setAddress(addresss[ThreadLocalRandom.current().nextInt(3)]);
			user.setGender((rand & 1) == 0 ? "girl" : "boy");
			user.setAge(ages[ThreadLocalRandom.current().nextInt(3)]);
			users.add(user);
		}
		return users;
	}

}

测试结果(查询2w次)：

数据量（users.size()）	并发数	平均查询时间
20w	10	1ms
50w	20	3ms
100w	50	9ms

测试机为thinkpad x240 i5 8g内存

4.总结

优点：
通过测试发现随着数据量的增大和并发数的提高，平均耗时并没有明显升高，并且响应时间都在10ms以内

缺点：

1. 不适合数据量太大的情况,因为需要把数据全部加载进内存
2. 不适合复杂查询
3. 不适合对name,id等唯一值做查询

后记

因为我们的查询业务比较简单，唯一的要求是速度，并且数据量也不大，每张表的数据量都不超过100w，所以使用这种方式比较合适。
在本篇文章中只谈到了如何创建索引，以及最基本的查询，在下一篇中我会继续说明如何更新索引，以及一些复杂查询，比如<,>,between and。

转载请注明出处