這是幀動畫系列的第一篇,系列文章目錄以下:java
Android 提供了AnimationDrawable
用於實現幀動畫。在動畫開始以前,全部幀的圖片都被解析並佔用內存,一旦動畫較複雜幀數較多,在低配置手機上容易發生 OOM。即便不發生 OOM,也會對內存形成不小的壓力。下面代碼展現了一個幀數爲4的幀動畫:git
AnimationDrawable drawable = new AnimationDrawable();
drawable.addFrame(getDrawable(R.drawable.frame1), frameDuration);
drawable.addFrame(getDrawable(R.drawable.frame2), frameDuration);
drawable.addFrame(getDrawable(R.drawable.frame3), frameDuration);
drawable.addFrame(getDrawable(R.drawable.frame4), frameDuration);
drawable.setOneShot(true);
ImageView ivFrameAnim = ((ImageView) findViewById(R.id.frame_anim));
ivFrameAnim.setImageDrawable(drawable);
drawable.start();
複製代碼
有沒有什麼辦法讓幀動畫的數據逐幀加載,而不是一次性所有加載到內存?SurfaceView
就提供了這種能力。github
屏幕的顯示機制和幀動畫相似,也是一幀一幀的連環畫,只不過刷新頻率很高,感受像連續的。爲了顯示一幀,須要經歷計算和渲染兩個過程,CPU 先計算出這一幀的圖像數據並寫入內存,而後調用 OpenGL 命令將內存中數據渲染成圖像存放在 GPU Buffer 中,顯示設備每隔必定時間從 Buffer 中獲取圖像並顯示。算法
上述過程當中的計算,對於View
來講,就比如在主線程遍歷 View樹 以決定視圖畫多大(measure),畫在哪(layout),畫些啥(draw),計算結果存放在內存中,SurfaceFlinger 會調用 OpenGL 命令將內存中的數據渲染成圖像存放在 GPU Buffer 中。每隔16.6ms,顯示器從 Buffer 中取出幀並顯示。因此自定義 View 能夠經過重載onMeasure()
、onLayout()
、onDraw()
來定義幀內容,但不能定義幀刷新頻率。canvas
SurfaceView
能夠突破這個限制。並且它能夠將計算幀數據放到獨立的線程中進行。下面是自定義SurfaceView
的模版代碼:bash
public abstract class BaseSurfaceView extends SurfaceView implements SurfaceHolder.Callback {
public static final int DEFAULT_FRAME_DURATION_MILLISECOND = 50;
//用於計算幀數據的線程
private HandlerThread handlerThread;
private Handler handler;
//幀刷新頻率
private int frameDuration = DEFAULT_FRAME_DURATION_MILLISECOND;
//用於繪製幀的畫布
private Canvas canvas;
private boolean isAlive;
public BaseSurfaceView(Context context) {
super(context);
init();
}
protected void init() {
getHolder().addCallback(this);
//設置透明背景,不然SurfaceView背景是黑的
setBackgroundTransparent();
}
private void setBackgroundTransparent() {
getHolder().setFormat(PixelFormat.TRANSLUCENT);
setZOrderOnTop(true);
}
@Override
public void surfaceCreated(SurfaceHolder holder) {
isAlive = true;
startDrawThread();
}
@Override
public void surfaceChanged(SurfaceHolder holder, int format, int width, int height) {
}
@Override
public void surfaceDestroyed(SurfaceHolder holder) {
stopDrawThread();
isAlive = false;
}
//中止幀繪製線程
private void stopDrawThread() {
handlerThread.quit();
handler = null;
}
//啓動幀繪製線程
private void startDrawThread() {
handlerThread = new HandlerThread("SurfaceViewThread");
handlerThread.start();
handler = new Handler(handlerThread.getLooper());
handler.post(new DrawRunnable());
}
private class DrawRunnable implements Runnable {
@Override
public void run() {
if (!isAlive) {
return;
}
try {
//1.獲取畫布
canvas = getHolder().lockCanvas();
//2.繪製一幀
onFrameDraw(canvas);
} catch (Exception e) {
e.printStackTrace();
} finally {
//3.將幀數據提交
getHolder().unlockCanvasAndPost(canvas);
//4.一幀繪製結束
onFrameDrawFinish();
}
//不停的將本身推送到繪製線程的消息隊列以實現幀刷新
handler.postDelayed(this, frameDuration);
}
}
protected abstract void onFrameDrawFinish();
protected abstract void onFrameDraw(Canvas canvas);
}
複製代碼
HandlerThread
做爲獨立幀繪製線程,好處是能夠經過與其綁定的Handler
方便地實現「每隔一段時間刷新」,並且在Surface
被銷燬的時候能夠方便的調用HandlerThread.quit()
來結束線程執行的邏輯。DrawRunnable.run()
運用模版方法模式定義了繪製算法框架,其中幀繪製邏輯的具體實現被定義成兩個抽象方法,推遲到子類中實現,由於繪製的東西是多樣的,對於本文來講,繪製的就是一張張圖片,因此新建BaseSurfaceView
的子類FrameSurfaceView
:public class FrameSurfaceView extends BaseSurfaceView {
public static final int INVALID_BITMAP_INDEX = Integer.MAX_VALUE;
private List<Integer> bitmaps = new ArrayList<>();
//幀圖片
private Bitmap frameBitmap;
//幀索引
private int bitmapIndex = INVALID_BITMAP_INDEX;
private Paint paint = new Paint();
private BitmapFactory.Options options = new BitmapFactory.Options();
//幀圖片原始大小
private Rect srcRect;
//幀圖片目標大小
private Rect dstRect = new Rect();
private int defaultWidth;
private int defaultHeight;
public void setDuration(int duration) {
int frameDuration = duration / bitmaps.size();
setFrameDuration(frameDuration);
}
public void setBitmaps(List<Integer> bitmaps) {
if (bitmaps == null || bitmaps.size() == 0) {
return;
}
this.bitmaps = bitmaps;
//默認狀況下,計算第一幀圖片的原始大小
getBitmapDimension(bitmaps.get(0));
}
private void getBitmapDimension(Integer integer) {
final BitmapFactory.Options options = new BitmapFactory.Options();
options.inJustDecodeBounds = true;
BitmapFactory.decodeResource(this.getResources(), integer, options);
defaultWidth = options.outWidth;
defaultHeight = options.outHeight;
srcRect = new Rect(0, 0, defaultWidth, defaultHeight);
requestLayout();
}
public FrameSurfaceView(Context context) {
super(context);
}
@Override
protected void onLayout(boolean changed, int left, int top, int right, int bottom) {
super.onLayout(changed, left, top, right, bottom);
dstRect.set(0, 0, getWidth(), getHeight());
}
@Override
protected void onFrameDrawFinish() {
//在一幀繪製完後,直接回收它
recycleOneFrame();
}
//回收幀
private void recycleOneFrame() {
if (frameBitmap != null) {
frameBitmap.recycle();
frameBitmap = null;
}
}
@Override
protected void onFrameDraw(Canvas canvas) {
//繪製一幀前須要先清畫布,不然全部幀都疊在一塊兒同時顯示
clearCanvas(canvas);
if (!isStart()) {
return;
}
if (!isFinish()) {
drawOneFrame(canvas);
} else {
onFrameAnimationEnd();
}
}
//繪製一幀,是張Bitmap
private void drawOneFrame(Canvas canvas) {
frameBitmap = BitmapUtil.decodeOriginBitmap(getResources(), bitmaps.get(bitmapIndex), options);
canvas.drawBitmap(frameBitmap, srcRect, dstRect, paint);
bitmapIndex++;
}
private void onFrameAnimationEnd() {
reset();
}
private void reset() {
bitmapIndex = INVALID_BITMAP_INDEX;
}
//幀動畫是否結束
private boolean isFinish() {
return bitmapIndex >= bitmaps.size();
}
//幀動畫是否開始
private boolean isStart() {
return bitmapIndex != INVALID_BITMAP_INDEX;
}
//開始播放幀動畫
public void start() {
bitmapIndex = 0;
}
private void clearCanvas(Canvas canvas) {
paint.setXfermode(new PorterDuffXfermode(PorterDuff.Mode.CLEAR));
canvas.drawPaint(paint);
paint.setXfermode(new PorterDuffXfermode(PorterDuff.Mode.SRC));
}
}
複製代碼
FrameSurfaceView
繼承自BaseSurfaceView
,因此它複用了基類的繪製框架算法,而且定了本身每一幀的繪製內容:一張Bitmap
。Bitmap
資源 id 經過setBitmaps()
傳遞進來, 繪製一幀解析一張 ,在每一幀繪製完畢後,調用Bitmap.recycle()
釋放圖片 native 內存並去除 java 堆中圖片像素數據的引用。這樣當 GC 發生時,圖片像素數據能夠及時被回收。一切都是這麼地可以自圓其說,我火燒眉毛地運行代碼並打開AndroidStudio
的Profiler
標籤頁,切換到MEMORY
,想用真實內存數據驗證下性能。但殘酷的事實狠狠地打了下臉。。。屢次播放幀動畫後,內存佔用竟然比原生AnimationDrawable
還大,並且每播放一次,內存中都會多出 N 個Bitmap
對象(N爲幀動畫總幀數)。惟一使人欣慰的是,手動觸發 GC 後幀動畫圖片可以被回收。(AnimationDrawable
中的圖片數據不會被 GC)框架
緣由就在於自做聰明地及時回收,每一幀繪製完後幀數據被回收,那下一幀解析Bitmap
時只能新申請一塊內存。幀動畫每張圖片大小是一致的,是否是能複用上一幀Bitmap
的內存空間?因而乎有了下面這個版本的FrameSurfaceView
:dom
public class FrameSurfaceView extends BaseSurfaceView {
public static final int INVALID_BITMAP_INDEX = Integer.MAX_VALUE;
private List<Integer> bitmaps = new ArrayList<>();
private Bitmap frameBitmap;
private int bitmapIndex = INVALID_BITMAP_INDEX;
private Paint paint = new Paint();
private BitmapFactory.Options options;
private Rect srcRect;
private Rect dstRect = new Rect();
public void setDuration(int duration) {
int frameDuration = duration / bitmaps.size();
setFrameDuration(frameDuration);
}
public void setBitmaps(List<Integer> bitmaps) {
if (bitmaps == null || bitmaps.size() == 0) {
return;
}
this.bitmaps = bitmaps;
getBitmapDimension(bitmaps.get(0));
}
private void getBitmapDimension(Integer integer) {
final BitmapFactory.Options options = new BitmapFactory.Options();
options.inJustDecodeBounds = true;
BitmapFactory.decodeResource(this.getResources(), integer, options);
defaultWidth = options.outWidth;
defaultHeight = options.outHeight;
srcRect = new Rect(0, 0, defaultWidth, defaultHeight);;
}
public FrameSurfaceView(Context context) {
super(context);
}
@Override
protected void init() {
super.init();
//定義解析Bitmap參數爲可變類型,這樣才能複用Bitmap
options = new BitmapFactory.Options();
options.inMutable = true;
}
@Override
protected void onLayout(boolean changed, int left, int top, int right, int bottom) {
super.onLayout(changed, left, top, right, bottom);
dstRect.set(0, 0, getWidth(), getHeight());
}
@Override
protected int getDefaultWidth() {
return defaultWidth;
}
@Override
protected int getDefaultHeight() {
return defaultHeight;
}
@Override
protected void onFrameDrawFinish() {
//每幀繪製完畢後再也不回收
// recycle();
}
public void recycle() {
if (frameBitmap != null) {
frameBitmap.recycle();
frameBitmap = null;
}
}
@Override
protected void onFrameDraw(Canvas canvas) {
clearCanvas(canvas);
if (!isStart()) {
return;
}
if (!isFinish()) {
drawOneFrame(canvas);
} else {
onFrameAnimationEnd();
}
}
private void drawOneFrame(Canvas canvas) {
frameBitmap = BitmapUtil.decodeOriginBitmap(getResources(), bitmaps.get(bitmapIndex), options);
//複用上一幀Bitmap的內存
options.inBitmap = frameBitmap;
canvas.drawBitmap(frameBitmap, srcRect, dstRect, paint);
bitmapIndex++;
}
private void onFrameAnimationEnd() {
reset();
}
private void reset() {
bitmapIndex = INVALID_BITMAP_INDEX;
}
private boolean isFinish() {
return bitmapIndex >= bitmaps.size();
}
private boolean isStart() {
return bitmapIndex != INVALID_BITMAP_INDEX;
}
public void start() {
bitmapIndex = 0;
}
private void clearCanvas(Canvas canvas) {
paint.setXfermode(new PorterDuffXfermode(PorterDuff.Mode.CLEAR));
canvas.drawPaint(paint);
paint.setXfermode(new PorterDuffXfermode(PorterDuff.Mode.SRC));
}
}
複製代碼
Bitmap
的解析參數inBitmap
設置爲已經成功解析的Bitmap
對象以實現複用。這一次無論從新播放多少次幀動畫,內存中Bitmap
數量只會增長1,由於只在解析第一張圖片是分配了內存。而這塊內存能夠在FrameSurfaceView
生命週期結束時手動調用recycle()
回收。ide
爲了更清晰的展現,上述代碼段省略了一些和主題無關的自定義 View 細節,完整的代碼能夠點擊這裏。oop