Android平台Gallery2应用分析(七)---PhotoPage图片解码

PhotoPage图片解码
从前文可知，PhotoPage的图片解码始于PhotoPage的onResume()调用updateImageRequests()。先看下代码：

private void updateImageRequests() {
……
int currentIndex = mCurrentIndex;
MediaItem item = mData[currentIndex % DATA_CACHE_SIZE];
……
// 1. 遍历sImageFetchSeq，查看当前图片符合哪种类型，调用startTaskIfNeeded
Future<?> task = null;
for (int i = 0; i < sImageFetchSeq.length; i++) {
int offset = sImageFetchSeq[i].indexOffset;
int bit = sImageFetchSeq[i].imageBit;
if (bit == BIT_FULL_IMAGE && !mNeedFullImage) continue;
task = startTaskIfNeeded(currentIndex + offset, bit);
if (task != null) break;
}

// 2. 释放任务和内存
for (ImageEntry entry : mImageCache.values()) {
if (entry.screenNailTask != null && entry.screenNailTask != task) {
entry.screenNailTask.cancel();
entry.screenNailTask = null;
entry.requestedScreenNail = MediaObject.INVALID_DATA_VERSION;
}
if (entry.fullImageTask != null && entry.fullImageTask != task) {
entry.fullImageTask.cancel();
entry.fullImageTask = null;
entry.requestedFullImage = MediaObject.INVALID_DATA_VERSION;
}
}
}
接下来，重点分析startTaskIfNeeded()，看它是如何对一张图片做解析的。还是先看下面代码：

private Future<?> startTaskIfNeeded(int index, int which) {
if (index < mActiveStart || index >= mActiveEnd) return null;

ImageEntry entry = mImageCache.get(getPath(index));
if (entry == null) return null;
// 先得到当前图片，类型为LocalImage
MediaItem item = mData[index % DATA_CACHE_SIZE];
Utils.assertTrue(item != null);
long version = item.getDataVersion();

// 第一次代码执行暂时screenNailTask和fullImageTask为null
if (which == BIT_SCREEN_NAIL && entry.screenNailTask != null
&& entry.requestedScreenNail == version) {
return entry.screenNailTask;
} else if (which == BIT_FULL_IMAGE && entry.fullImageTask != null
&& entry.requestedFullImage == version) {
return entry.fullImageTask;
}
// 匹配到格式后，先创建ScreenNailJob、ScreenNailListener并返回screenNailTask
if (which == BIT_SCREEN_NAIL && entry.requestedScreenNail != version) {
entry.requestedScreenNail = version;
entry.screenNailTask = mThreadPool.submit(
new ScreenNailJob(item),
new ScreenNailListener(item));
// request screen nail
return entry.screenNailTask;
}
// 匹配到格式后，先创建FullImageJob、FullImageListener并返回fullImageTask
if (which == BIT_FULL_IMAGE && entry.requestedFullImage != version
&& (item.getSupportedOperations()
& MediaItem.SUPPORT_FULL_IMAGE) != 0) {
entry.requestedFullImage = version;
entry.fullImageTask = mThreadPool.submit(
new FullImageJob(item),
new FullImageListener(item));
// request full image
return entry.fullImageTask;
}
return null;
}

参数which就是静态数组sImageFetchSeq中的图片类型，android原生代码默认两种BIT_SCREEN_NAIL和BIT_FULL_IMAGE。当然我们也可以自己加入一种解析图片的格式，例如BIT_GIF_IMAGE。在详细分析下面的代码后，详细你自己加入一种图片格式应该问题不大。流程大致如下：

1） 从传入参数index获取当前图片MediaItem，图片为LocalImage类型。第一次执行时screenNailTask和fullImageTask为null，匹配到格式后，创建ScreenNailJob、ScreenNailListener（或者screenNailTask和fullImageTask为null），并最终返回screenNailTask或者fullImageTask。

2） 以FullImage为例。ThreadPool的机制这里再大致讲述一下。看下mThreadPool.submit的代码：

public <T> Future<T> submit(Job<T> job, FutureListener<T> listener) {
Worker<T> w = new Worker<T>(job, listener);
mExecutor.execute(w);
return w;
}

由前文分析，execute会启动Worker的线程进入run()函数：

@Override
public void run() {
T result = null;

// A job is in CPU mode by default. setMode returns false
// if the job is cancelled.
if (setMode(MODE_CPU)) {
try {
// 1、执行job的run()
result = mJob.run(this);
} catch (Throwable ex) {
Log.w(TAG, "Exception in running a job", ex);
}
}

synchronized(this) {
setMode(MODE_NONE);
mResult = result;
mIsDone = true;
notifyAll();
}
// 2、执行完毕，调用listener的onFutureDone
if (mListener != null) mListener.onFutureDone(this);
}

这里面分两步：

        2.1）执行job的run()。这里会调用传入参数new FullImageJob (item).run()。

private class FullImageJob implements Job<BitmapRegionDecoder> {
private MediaItem mItem;

public FullImageJob(MediaItem item) {
mItem = item;
}

@Override
public BitmapRegionDecoder run(JobContext jc) {
if (isTemporaryItem(mItem)) {
return null;
}
return mItem.requestLargeImage().run(jc);
}
}

这段代码实际就是在线程池的某个线程中执行LocalImage的requestLargeImage().run(jc)。而该函数会创建一个LocalLargeImageRequest对象，而run(jc)实际就是DecodeUtils.createBitmapRegionDecoder(jc, mLocalFilePath, false)，最终创建一个BitmapRegionDecoder实例。该实例会调用JNI层的BitmapRegionDecoder中的nativeNewInstanceFromStream接口，在doBuildTileIndex(JNIEnv*
env, SkStream* stream)里可以看到，图片会根据stream的header判断解码器是SkJPEGImageDecoder，SkPNGImageDecoder，SkBMPImageDecoder还是SkWEBPImageDecoder等，最后得到解码数据。

        2.2） 调用FullImageListener的onFutureDone。此时会将2.1)中创建好的BitmapRegionDecoder实例返回传给mFuture。而FullImageListener则再发送一个MSG_RUN_OBJECT消息给MainThread，MainThread再执行FullImageListener的run()，即再执行updateFullImage(mPath, mFuture)。

private void updateFullImage(Path path, Future<BitmapRegionDecoder> future) {
ImageEntry entry = mImageCache.get(path);
……
entry.fullImageTask = null;
entry.fullImage = future.get();
if (entry.fullImage != null) {
if (path == getPath(mCurrentIndex)) {
updateTileProvider(entry);
mPhotoView.notifyImageChange(0);
}
}
updateImageRequests();
}
其中mPhotoView.notifyImageChange(0)取到当前图片后并reload，其中mPictures当前为FullPicture。

public void notifyImageChange(int index) {
……
mPictures.get(index).reload();
……
invalidate();
}

那么再看看FullPicture的reload(), 该函数会对mTileView的screenNail做更新。

@Override
public void reload() {
// mImageWidth and mImageHeight will get updated
mTileView.notifyModelInvalidated();
……
setScreenNail(mModel.getScreenNail(0));
……
}

代码段中的mModel就是PhotoDataAdapter。mModel.getScreenNail(0)得到当前图片的ScreenNail以做更新。由此可知，我们看到的全屏的图片，就是TileImageView类型的。

欢迎转载和技术交流，转载请帮忙注明出处，http://blog.csdn.net/discovery_by_joseph，谢谢！