1 概述
Okio是一個(gè)對(duì)java.io和java.nio進(jìn)行補(bǔ)充的庫(kù),使數(shù)據(jù)訪問(wèn),保存和處理變得更容易。
Okio的主要功能是圍繞著B(niǎo)yteString 和 Buffer 兩個(gè)類展開(kāi)的:
1> ByteString是一個(gè)immutable的字節(jié)序列。在java中,String代表的是字符串,ByteString和String很相似,只不過(guò)是用來(lái)處理字節(jié)串的,同時(shí)也提供了常用的操作,比如對(duì)數(shù)據(jù)進(jìn)行十六進(jìn)制(hex)、base64 和 UTF-8 格式的編碼和解碼,equals、substring等操作。
2> Buffer是一個(gè)mutable的字節(jié)序列。 和ArrayList類似,不需要提前設(shè)置緩沖區(qū)大小。讀取數(shù)據(jù)和寫入數(shù)據(jù)和隊(duì)列類似,從它的head讀取數(shù)據(jù),往它的tail寫入數(shù)據(jù),而且不用考慮容量、位置等因素。
java.io設(shè)計(jì)的一個(gè)優(yōu)雅部分是如何將stream分層以進(jìn)行加密和壓縮等轉(zhuǎn)換。 Okio包括自己的stream類型,稱為Source和Sink,和InputStream和OutputStream的工作方式類似,但有一些關(guān)鍵的區(qū)別:
1> Timeouts: 提供對(duì)底層I/O訪問(wèn)的超時(shí)機(jī)制。
2> Source和Sink的API非常簡(jiǎn)潔,易于實(shí)現(xiàn)。
3> 雖然Source和Sink的只提供了三個(gè)方法,但是BufferedSource和BufferedSink接口提供了更豐富的方法(比如針對(duì)不同類型的read和write方法),以應(yīng)對(duì)更加復(fù)雜的場(chǎng)景。
4> 不在區(qū)分byte stream和char stream,它們都是數(shù)據(jù),可以按照任意類型進(jìn)行讀寫。
2 Segment和SegmentPool
Segment的源碼不到200行,直接通過(guò)源碼來(lái)理解Segment的實(shí)現(xiàn)原理也是很簡(jiǎn)單的,首先來(lái)看一下Segment中的所有的字段:
/** Segment可以保存的最大字節(jié)數(shù) */
static final int SIZE = 8192;
/** Segment被共享時(shí)最小的字節(jié)數(shù) */
static final int SHARE_MINIMUM = 1024;
/** Segment中保存數(shù)據(jù)的字節(jié)數(shù)組 */
final byte[] data;
/** 字節(jié)數(shù)組data中被當(dāng)前Segment實(shí)例使用的區(qū)間的第一個(gè)字節(jié)的下標(biāo) */
int pos;
/** 字節(jié)數(shù)組data中被當(dāng)前Segment實(shí)例使用的區(qū)間之后的第一個(gè)字節(jié)的下標(biāo) */
int limit;
/** 代表字節(jié)數(shù)組data是否被 >=2 個(gè)Segment實(shí)例共用*/
boolean shared;
/** 代表字節(jié)數(shù)組data中最后一段被使用的區(qū)間是不是被當(dāng)前Segment實(shí)例占有*/
boolean owner;
/** 當(dāng)前Segment實(shí)例的后置節(jié)點(diǎn) */
Segment next;
/** 當(dāng)前Segment實(shí)例的前置節(jié)點(diǎn) */
Segment prev;
shared、owner的作用:
在向Segment中寫入數(shù)據(jù)時(shí),首先用owner判斷當(dāng)前Segment實(shí)例對(duì)應(yīng)的數(shù)據(jù)區(qū)間(字節(jié)數(shù)組data被使用的區(qū)間)之后是否可以寫入數(shù)據(jù),接著用shared判斷當(dāng)前Segment實(shí)例對(duì)應(yīng)的數(shù)據(jù)區(qū)間之前是否可以寫入數(shù)據(jù),體現(xiàn)在了Segment的writeTo方法中。
接下來(lái)依次分析Segment中的方法:
/**
* 從循環(huán)雙向鏈表中移除當(dāng)前Segment實(shí)例,返回當(dāng)前Segment實(shí)例的后置節(jié)點(diǎn)。
*/
public @Nullable Segment pop() {
Segment result = next != this ? next : null;
prev.next = next;
next.prev = prev;
next = null;
prev = null;
return result;
}
/**
* 在循環(huán)雙向鏈表中的當(dāng)前Segment實(shí)例之后插入segment實(shí)例,返回被插入的segment實(shí)例。
*/
public Segment push(Segment segment) {
segment.prev = this;
segment.next = next;
next.prev = segment;
next = segment;
return segment;
}
上面的兩個(gè)方法相信大家一看就明白了,就不再贅敘了。
/**
* 將當(dāng)前Segment實(shí)例中的字節(jié)數(shù)組data進(jìn)行分割,從而得到兩個(gè)Segment實(shí)例.
* 字節(jié)數(shù)組data中[pos..pos+byteCount)區(qū)間的數(shù)據(jù)屬于第一個(gè)segment.
* [pos+byteCount..limit)區(qū)間的數(shù)據(jù)屬于第二個(gè)segment.
*
* 返回第一個(gè)segment.
*/
public Segment split(int byteCount) {
if (byteCount <= 0 || byteCount > limit - pos) throw new IllegalArgumentException();
Segment prefix;
// We have two competing performance goals:
// - Avoid copying data. We accomplish this by sharing segments.
// - Avoid short shared segments. These are bad for performance because they are readonly and
// may lead to long chains of short segments.
// To balance these goals we only share segments when the copy will be large.
if (byteCount >= SHARE_MINIMUM) {
prefix = new Segment(this);
} else {
prefix = SegmentPool.take();
System.arraycopy(data, pos, prefix.data, 0, byteCount);
}
prefix.limit = prefix.pos + byteCount;
pos += byteCount;
prev.push(prefix);
return prefix;
}
/**
* 當(dāng)當(dāng)前Segment實(shí)例的前置節(jié)點(diǎn)中的空閑空間可以容納當(dāng)前Segment實(shí)例中的數(shù)據(jù).
* 則將當(dāng)前Segment實(shí)例中的數(shù)據(jù)拷貝到前置節(jié)點(diǎn)中并且將當(dāng)前Segment實(shí)例回收到SegmentPool中。
*/
public void compact() {
if (prev == this) throw new IllegalStateException();
if (!prev.owner) return; // Cannot compact: prev isn't writable.
int byteCount = limit - pos;
int availableByteCount = SIZE - prev.limit + (prev.shared ? 0 : prev.pos);
if (byteCount > availableByteCount) return; // Cannot compact: not enough writable space.
writeTo(prev, byteCount);
pop();
SegmentPool.recycle(this);
}
/** 將當(dāng)前Segment實(shí)例中的前byteCount個(gè)字節(jié)的數(shù)據(jù)復(fù)制放到sink中 */
public void writeTo(Segment sink, int byteCount) {
if (!sink.owner) throw new IllegalArgumentException();
if (sink.limit + byteCount > SIZE) {
// We can't fit byteCount bytes at the sink's current position. Shift sink first.
if (sink.shared) throw new IllegalArgumentException();
if (sink.limit + byteCount - sink.pos > SIZE) throw new IllegalArgumentException();
System.arraycopy(sink.data, sink.pos, sink.data, 0, sink.limit - sink.pos);
sink.limit -= sink.pos;
sink.pos = 0;
}
System.arraycopy(data, pos, sink.data, sink.limit, byteCount);
sink.limit += byteCount;
pos += byteCount;
}
上面的注釋已經(jīng)非常清晰了,這里就不再解釋了。
上面的compact方法中用到了SegmentPool.recycle(this)來(lái)回收Segment實(shí)例,那下面就來(lái)講解SegmentPool類,該類的存在就是為了避免GC churn(高頻率的創(chuàng)建和回收Segment實(shí)例會(huì)導(dǎo)致GC churn)和zero-fill(創(chuàng)建Segment實(shí)例時(shí)字節(jié)數(shù)組data需要zero-fill),SegmentPool實(shí)例中用一個(gè)單向的鏈表來(lái)保存回收的Segment實(shí)例,首先來(lái)看看Segment的源代碼:
/**
* 用于保存被回收的Segment實(shí)例,該類的存在就是為了避免GC churn和zero-fill
* SegmentPool實(shí)例是線程安全的靜態(tài)單例
*/
final class SegmentPool {
/** SegmentPool實(shí)例中保存的最大字節(jié)數(shù),因此SegmentPool中最多保存8個(gè)Segment實(shí)例 */
// TODO: Is 64 KiB a good maximum size? Do we ever have that many idle segments?
static final long MAX_SIZE = 64 * 1024; // 64 KiB.
/** SegmentPool實(shí)例中是通過(guò)單向非循環(huán)的鏈表來(lái)保存數(shù)據(jù)的,next代表鏈表中的第一個(gè)Segment實(shí)例 */
static @Nullable Segment next;
/** SegmentPool實(shí)例中的字節(jié)總數(shù). */
static long byteCount;
private SegmentPool() {
}
static Segment take() {
synchronized (SegmentPool.class) {
if (next != null) {
Segment result = next;
next = result.next;
result.next = null;
byteCount -= Segment.SIZE;
return result;
}
}
return new Segment(); // Pool is empty. Don't zero-fill while holding a lock.
}
static void recycle(Segment segment) {
if (segment.next != null || segment.prev != null) throw new IllegalArgumentException();
if (segment.shared) return; // This segment cannot be recycled.
synchronized (SegmentPool.class) {
if (byteCount + Segment.SIZE > MAX_SIZE) return; // Pool is full.
byteCount += Segment.SIZE;
segment.next = next;
segment.pos = segment.limit = 0;
next = segment;
}
}
}
是不是很簡(jiǎn)單,一共也不到70行,一共提供了兩個(gè)方法:
Segment take():從SegmentPool實(shí)例中獲取被回收的Segment實(shí)例,如果SegmentPool實(shí)例是空的,則創(chuàng)建一個(gè)Segment實(shí)例返回。
void recycle(Segment segment):回收segment實(shí)例。
3 Buffer
Buffer內(nèi)部使用Segment的雙向鏈表來(lái)保存數(shù)據(jù),Segment內(nèi)部使用字節(jié)數(shù)組保存數(shù)據(jù)。 將數(shù)據(jù)從一個(gè)Buffer移動(dòng)到另一個(gè)Buffer時(shí),會(huì)通過(guò)轉(zhuǎn)讓Segment的所有權(quán),而不用拷貝數(shù)據(jù),從而節(jié)省性能上的開(kāi)銷。下面通過(guò)一張圖來(lái)描述一下Buffer中雙向循環(huán)鏈表和SegmentPool單向非循環(huán)鏈表:

下面通過(guò)一張類圖來(lái)整體的描述一下Buffer:

為了更加清晰的理解上圖,就需要簡(jiǎn)單的了解一下裝飾者模式:
1> 定義:Attach additional responsibilities to an object dynamically keeping the same interface.Decorators provide a flexible alternative to subclassing for extending functionality. (動(dòng)態(tài)的給一個(gè)對(duì)象添加額外的職責(zé)。就增加功能來(lái)說(shuō),裝飾者模式相比生成子類更加靈活。)
2> 裝飾者模式通用類圖

說(shuō)明一下類圖中的四個(gè)角色:
Component抽象組件:Component是一個(gè)接口或者是抽象類,在裝飾者模式中,必然有一個(gè)最基本、最核心、最原始的接口或抽象類充當(dāng)Component抽象組件。對(duì)應(yīng)于在Okio框架中的BufferedSource和BufferedSink接口。
ConcreteComponent具體組件:對(duì)Component抽象組件的實(shí)現(xiàn),將要被裝飾的類。對(duì)應(yīng)于Okio框架中的Buffer。
Decorator裝飾者:一般是一個(gè)抽象類,繼承至Component抽象組件。一定擁有一個(gè)指向Component抽象組件的priavte字段。
ConcreteDecorator具體裝飾者:對(duì)Decorator裝飾者的實(shí)現(xiàn),用來(lái)裝飾ConcreteComponent具體組件。
在Okio框架中沒(méi)有細(xì)分Decorator和ConcreteDecorator,只有兩個(gè)具體裝飾類RealBufferedSource和RealBufferedSink。
Source是用來(lái)對(duì)數(shù)據(jù)來(lái)源的封裝,Sink是對(duì)數(shù)據(jù)消費(fèi)的封裝,在Okio工具類中,為Source提供了四種數(shù)據(jù)來(lái)源:Socket、InputStream、File和Path,同樣為Sink提供了四種數(shù)據(jù)消費(fèi):Socket、OutputStream、File和Path,接下來(lái)針對(duì)Socket舉例分析:
public void testSocket(Socket socket) {
try {
Source source = Okio.source(socket);
BufferedSource bufferedSource = Okio.buffer(source);
bufferedSource.timeout().timeout(500, TimeUnit.MILLISECONDS);
String data = bufferedSource.readString(Charset.forName("UTF-8"));
} catch (IOException e) {
e.printStackTrace();
}
}
上面就是使用Okio框架讀數(shù)據(jù)的過(guò)程,下面我們就來(lái)看看源碼中是如何實(shí)現(xiàn)的:
// BufferedSource的方法:
@Override public String readString(Charset charset) throws IOException {
if (charset == null) throw new IllegalArgumentException("charset == null");
buffer.writeAll(source);
return buffer.readString(charset);
}
// Buffered的方法:
@Override public long writeAll(Source source) throws IOException {
if (source == null) throw new IllegalArgumentException("source == null");
long totalBytesRead = 0;
for (long readCount; (readCount = source.read(this, Segment.SIZE)) != -1; ) {
totalBytesRead += readCount;
}
return totalBytesRead;
}
@Override public String readString(Charset charset) {
try {
return readString(size, charset);
} catch (EOFException e) {
throw new AssertionError(e);
}
}
@Override public String readString(long byteCount, Charset charset) throws EOFException {
checkOffsetAndCount(size, 0, byteCount);
if (charset == null) throw new IllegalArgumentException("charset == null");
if (byteCount > Integer.MAX_VALUE) {
throw new IllegalArgumentException("byteCount > Integer.MAX_VALUE: " + byteCount);
}
if (byteCount == 0) return "";
Segment s = head;
if (s.pos + byteCount > s.limit) {
// If the string spans multiple segments, delegate to readBytes().
return new String(readByteArray(byteCount), charset);
}
String result = new String(s.data, s.pos, (int) byteCount, charset);
s.pos += byteCount;
size -= byteCount;
if (s.pos == s.limit) {
head = s.pop();
SegmentPool.recycle(s);
}
return result;
}
// Okio的方法:
/**
* Returns a new source that buffers reads from {@code source}. The returned
* source will perform bulk reads into its in-memory buffer. Use this wherever
* you read a source to get an ergonomic and efficient access to data.
*/
public static BufferedSource buffer(Source source) {
return new RealBufferedSource(source);
}
/**
* Returns a source that reads from {@code socket}. Prefer this over {@link
* #source(InputStream)} because this method honors timeouts. When the socket
* read times out, the socket is asynchronously closed by a watchdog thread.
*/
public static Source source(Socket socket) throws IOException {
if (socket == null) throw new IllegalArgumentException("socket == null");
AsyncTimeout timeout = timeout(socket);
Source source = source(socket.getInputStream(), timeout);
return timeout.source(source);
}
private static Source source(final InputStream in, final Timeout timeout) {
if (in == null) throw new IllegalArgumentException("in == null");
if (timeout == null) throw new IllegalArgumentException("timeout == null");
return new Source() {
@Override public long read(Buffer sink, long byteCount) throws IOException {
if (byteCount < 0) throw new IllegalArgumentException("byteCount < 0: " + byteCount);
if (byteCount == 0) return 0;
try {
timeout.throwIfReached();
Segment tail = sink.writableSegment(1);
int maxToCopy = (int) Math.min(byteCount, Segment.SIZE - tail.limit);
int bytesRead = in.read(tail.data, tail.limit, maxToCopy);
if (bytesRead == -1) return -1;
tail.limit += bytesRead;
sink.size += bytesRead;
return bytesRead;
} catch (AssertionError e) {
if (isAndroidGetsocknameError(e)) throw new IOException(e);
throw e;
}
}
@Override public void close() throws IOException {
in.close();
}
@Override public Timeout timeout() {
return timeout;
}
@Override public String toString() {
return "source(" + in + ")";
}
};
}
上面方法的流程可以概括如下:
1> 利用Source的public long read(Buffer sink, long byteCount)方法從Socket輸入流中讀取數(shù)據(jù)到Buffer實(shí)例中。
2> 接著調(diào)用Buffer的public String readString(Charset charset)方法將Buffer實(shí)例中的數(shù)據(jù)讀取到String對(duì)象中并且返回。
通過(guò)Okio框架寫數(shù)據(jù)的過(guò)程與讀數(shù)據(jù)的過(guò)程類似,只不過(guò)過(guò)程相反,就不再贅敘了。
下面給出Okio框架讀寫String數(shù)據(jù)的流程圖:

對(duì)于Okio框架讀寫其他類型數(shù)據(jù)也是類似的過(guò)程。
在上面的例子中還用到了TimeOut機(jī)制,其實(shí)Okio實(shí)現(xiàn)了兩種超時(shí)機(jī)制:
1> TimeOut 同步超時(shí)機(jī)制
利用throwIfReached方法在數(shù)據(jù)讀取過(guò)程中輪詢判斷是否超時(shí)。
2> AsyncTimeout 異步超時(shí)機(jī)制
由于通過(guò)Socket來(lái)讀寫數(shù)據(jù)會(huì)阻塞線程,所以用的是異步超時(shí)機(jī)制。
有興趣的同學(xué)可以自己閱讀源碼來(lái)分析超時(shí)機(jī)制。