Springboot 中RedisTemplate使用scan来获取所有的key底层做了哪些事情

直接上代码,围绕着代码来讨论

java 复制代码
redisTemplate.execute((RedisCallback<Object>) (connection) -> {
    Cursor<byte[]> scan = connection
            .scan(ScanOptions.scanOptions().count(2).match("*").build());
    scan.forEachRemaining((bytes) -> {
        System.out.println(new String(bytes));
    });
    //while (scan.hasNext()) {
    //    System.out.println(new String(scan.next()));
    //}
    return null;
});

通过上述代码我们可以看到,可以通过迭代器来获取到redis中单库中所有的key,但这样底层是怎么做的呢?我们现在假定有两种方案去实现这种方案。

① redis 客户端(jedis、lettuce)scan执行之后返回了所有的数据,迭代器next()每次都从数据集中取出一条消费。但这样有个疑点 => ScanOptions.scanOptions().count(2).match("*").build() 我们知道count(2)代表着每次只向redis中请求2条返回数据呀。如果全部返回也不符合scan命令的设计!

② 每次迭代next()都要判断是否数据集中还有数据,没有的话去redis中通过游标取下次一的数据集(2条)。然后将获取到数据集迭代器替换到游标中,上一个数据集回收(防止内存过大),使迭代器可以正常流转。

那上面两条都是我自己的猜测,可能两条都不对,也有可能两条对一条,具体我们还是要看源码怎么做的。

分析源码

Cursor<byte[]> scan = connection.scan(ScanOptions.scanOptions().count(2).match("*").build()); 这段代码返回了一个Cursor 游标,我们看看他的具体实现是这个类:org.springframework.data.redis.core.ScanCursor

这个类实现了Cursor,实现了迭代器的所有方法。

当我们执行forEachRemaining或者hasNext()的时候都会去判断

java 复制代码
@Override
public boolean hasNext() {

	assertCursorIsOpen();

	// delegate可以理解为数据集的迭代器,是一个list类型,从redis获取到的数据集会放到这个list中
	while (!delegate.hasNext() && !CursorState.FINISHED.equals(state)) {
		// 可以看到如果说数据集读取完了,会继续去往redis中取下一组
		scan(cursorId);
	}

	if (delegate.hasNext()) {
		return true;
	}

	return cursorId > 0;
}

scan 方法

java 复制代码
private void scan(long cursorId) {
	// 去请求redis获取下一组key
	ScanIteration<T> result = doScan(cursorId, this.scanOptions);
	// 将数据集
	processScanResult(result);
}

processScanResult 方法

java 复制代码
private void processScanResult(ScanIteration<T> result) {

	// redis 中的游标
	cursorId = result.getCursorId();

	if (isFinished(cursorId)) {
		state = CursorState.FINISHED;
	}

	if (!CollectionUtils.isEmpty(result.getItems())) {
		// 替换迭代器 *** 重要代码
		delegate = result.iterator();
	} else {
		resetDelegate();
	}
}

可以看到forEachRemaining其实也要去先判断数据集中有值。

自此我们验证了上述的两条推断中第二条是争取的,每次迭代器流转的时候都会去判断是否还有数据,没有的话就会从redis取到新的数据集,然后把数据集的迭代器替换到游标中,从而实现了redis单库取出所有的key。spring-data-redis中实现的很优雅,很巧妙,我们也可以学习这种设计模式来批量获取远程分页数据等。最后我们来看下ScanCursor的整体代码:

java 复制代码
/*
 * Copyright 2014-2022 the original author or authors.
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *      https://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */
package org.springframework.data.redis.core;

import java.util.Collections;
import java.util.Iterator;
import java.util.NoSuchElementException;

import org.springframework.dao.InvalidDataAccessApiUsageException;
import org.springframework.lang.Nullable;
import org.springframework.util.Assert;
import org.springframework.util.CollectionUtils;

/**
 * Redis client agnostic {@link Cursor} implementation continuously loading additional results from Redis server until
 * reaching its starting point {@code zero}. <br />
 * <strong>Note:</strong> Please note that the {@link ScanCursor} has to be initialized ({@link #open()} prior to usage.
 *
 * @author Christoph Strobl
 * @author Thomas Darimont
 * @author Duobiao Ou
 * @author Marl Paluch
 * @param <T>
 * @since 1.4
 */
public abstract class ScanCursor<T> implements Cursor<T> {

	private CursorState state; // 游标状态
	private long cursorId; // 游标当前位置
	private Iterator<T> delegate; // 从redis获取到的结果集的迭代器,只要实现过Iterator的类都可以被代理
	private final ScanOptions scanOptions; // 配置信息
	private long position; // 位置

	/**
	 * Crates new {@link ScanCursor} with {@code id=0} and {@link ScanOptions#NONE}
	 */
	public ScanCursor() {
		this(ScanOptions.NONE);
	}

	/**
	 * Crates new {@link ScanCursor} with {@code id=0}.
	 *
	 * @param options the scan options to apply.
	 */
	public ScanCursor(ScanOptions options) {
		this(0, options);
	}

	/**
	 * Crates new {@link ScanCursor} with {@link ScanOptions#NONE}
	 *
	 * @param cursorId the cursor Id.
	 */
	public ScanCursor(long cursorId) {
		this(cursorId, ScanOptions.NONE);
	}

	/**
	 * Crates new {@link ScanCursor}
	 *
	 * @param cursorId the cursor Id.
	 * @param options Defaulted to {@link ScanOptions#NONE} if {@code null}.
	 */
	public ScanCursor(long cursorId, @Nullable ScanOptions options) {

		this.scanOptions = options != null ? options : ScanOptions.NONE;
		this.cursorId = cursorId;
		this.state = CursorState.READY;
		this.delegate = Collections.emptyIterator();
	}

	private void scan(long cursorId) {

		ScanIteration<T> result = doScan(cursorId, this.scanOptions);
		processScanResult(result);
	}

	/**
	 * Performs the actual scan command using the native client implementation. The given {@literal options} are never
	 * {@code null}.
	 *
	 * @param cursorId
	 * @param options
	 * @return
	 */
	protected abstract ScanIteration<T> doScan(long cursorId, ScanOptions options);

	/**
	 * Initialize the {@link Cursor} prior to usage.
	 */
	public final ScanCursor<T> open() {

		if (!isReady()) {
			throw new InvalidDataAccessApiUsageException("Cursor already " + state + ". Cannot (re)open it.");
		}

		state = CursorState.OPEN;
		doOpen(cursorId);

		return this;
	}

	/**
	 * Customization hook when calling {@link #open()}.
	 *
	 * @param cursorId
	 */
	protected void doOpen(long cursorId) {
		scan(cursorId);
	}

	private void processScanResult(ScanIteration<T> result) {

		cursorId = result.getCursorId();

		if (isFinished(cursorId)) {
			state = CursorState.FINISHED;
		}

		if (!CollectionUtils.isEmpty(result.getItems())) {
			delegate = result.iterator();
		} else {
			resetDelegate();
		}
	}

	/**
	 * Check whether {@code cursorId} is finished.
	 *
	 * @param cursorId the cursor Id
	 * @return {@literal true} if the cursor is considered finished, {@literal false} otherwise.s
	 * @since 2.1
	 */
	protected boolean isFinished(long cursorId) {
		return cursorId == 0;
	}

	private void resetDelegate() {
		delegate = Collections.emptyIterator();
	}

	/*
	 * (non-Javadoc)
	 * @see org.springframework.data.redis.core.Cursor#getCursorId()
	 */
	@Override
	public long getCursorId() {
		return cursorId;
	}

	/*
	 * (non-Javadoc)
	 * @see java.util.Iterator#hasNext()
	 */
	@Override
	public boolean hasNext() {

		assertCursorIsOpen();

		while (!delegate.hasNext() && !CursorState.FINISHED.equals(state)) {
			scan(cursorId);
		}

		if (delegate.hasNext()) {
			return true;
		}

		return cursorId > 0;
	}

	private void assertCursorIsOpen() {

		if (isReady() || isClosed()) {
			throw new InvalidDataAccessApiUsageException("Cannot access closed cursor. Did you forget to call open()?");
		}
	}

	/*
	 * (non-Javadoc)
	 * @see java.util.Iterator#next()
	 */
	@Override
	public T next() {

		assertCursorIsOpen();

		if (!hasNext()) {
			throw new NoSuchElementException("No more elements available for cursor " + cursorId + ".");
		}

		T next = moveNext(delegate);
		position++;

		return next;
	}

	/**
	 * Fetch the next item from the underlying {@link Iterable}.
	 *
	 * @param source
	 * @return
	 */
	protected T moveNext(Iterator<T> source) {
		return source.next();
	}

	/*
	 * (non-Javadoc)
	 * @see java.util.Iterator#remove()
	 */
	@Override
	public void remove() {
		throw new UnsupportedOperationException("Remove is not supported");
	}

	/*
	 * (non-Javadoc)
	 * @see java.io.Closeable#close()
	 */
	@Override
	public final void close() {

		try {
			doClose();
		} finally {
			state = CursorState.CLOSED;
		}
	}

	/**
	 * Customization hook for cleaning up resources on when calling {@link #close()}.
	 */
	protected void doClose() {}

	/*
	 * (non-Javadoc)
	 * @see org.springframework.data.redis.core.Cursor#isClosed()
	 */
	@Override
	public boolean isClosed() {
		return state == CursorState.CLOSED;
	}

	protected final boolean isReady() {
		return state == CursorState.READY;
	}

	protected final boolean isOpen() {
		return state == CursorState.OPEN;
	}

	/*
	 * (non-Javadoc)
	 * @see org.springframework.data.redis.core.Cursor#getPosition()
	 */
	@Override
	public long getPosition() {
		return position;
	}

	/**
	 * @author Thomas Darimont
	 */
	enum CursorState {
		READY, OPEN, FINISHED, CLOSED;
	}
}
相关推荐
向前看-1 小时前
验证码机制
前端·后端
黄油饼卷咖喱鸡就味增汤拌孜然羊肉炒饭3 小时前
SpringBoot如何实现缓存预热?
java·spring boot·spring·缓存·程序员
超爱吃士力架3 小时前
邀请逻辑
java·linux·后端
AskHarries5 小时前
Spring Cloud OpenFeign快速入门demo
spring boot·后端
isolusion6 小时前
Springboot的创建方式
java·spring boot·后端
Yvemil76 小时前
《开启微服务之旅:Spring Boot Web开发举例》(一)
前端·spring boot·微服务
zjw_rp6 小时前
Spring-AOP
java·后端·spring·spring-aop
TodoCoder7 小时前
【编程思想】CopyOnWrite是如何解决高并发场景中的读写瓶颈?
java·后端·面试
凌虚8 小时前
Kubernetes APF(API 优先级和公平调度)简介
后端·程序员·kubernetes
星河梦瑾8 小时前
SpringBoot相关漏洞学习资料
java·经验分享·spring boot·安全