RecyclerView自定义StaggeredGridLayoutManager实现EPG布局

实现布局：

最近在做EPG功能，要求能够展示每个节目对应时间轴的预览节目信息，这种布局可以采用RecyclerView的瀑布流布局实现，但是测试的时候发现Item的position是没有规律的，这样就不能够很好的根据positoin去区分是哪个节目的EPG信息从而显示到不同行数上，所以需要从StaggeredGridLayoutManager源码改起

修改getNextSpan函数

/*** Finds the span for the next view.*/
private StaggeredGridLayoutManager.Span getNextSpan(LayoutState layoutState) {final boolean preferLastSpan = preferLastSpan(layoutState.mLayoutDirection);final int startIndex, endIndex, diff;if (preferLastSpan) {startIndex = mSpanCount - 1;endIndex = -1;diff = -1;} else {startIndex = 0;endIndex = mSpanCount;diff = 1;}//布局排列方向，表示从起始排到末尾if (layoutState.mLayoutDirection == LayoutState.LAYOUT_END) {Span min = null;int minLine = Integer.MAX_VALUE;final int defaultLine = mPrimaryOrientation.getStartAfterPadding();//for循环是找到布局中哪个列的最后一个Item的末尾处坐标最小for (int i = startIndex; i != endIndex; i += diff) {final Span other = mSpans[i];int otherLine = other.getEndLine(defaultLine);if (otherLine < minLine) {min = other;minLine = otherLine;}}return min;} else { //布局排列方向，表示从末尾排到起始Span max = null;int maxLine = Integer.MIN_VALUE;final int defaultLine = mPrimaryOrientation.getEndAfterPadding();//for循环是找到布局中哪个列的第一个Item的起始处坐标值最大for (int i = startIndex; i != endIndex; i += diff) {final Span other = mSpans[i];int otherLine = other.getStartLine(defaultLine);if (otherLine > maxLine) {max = other;maxLine = otherLine;}}return max;}
}

该函数是获取下一个Span，每次获取都会先计算当前哪个Span的末尾处最小，最小则返回该Span，以便下一个View添加到这个Span上，我们可以增加一个int变量，每次调用+1，循环去Span数组获取，修改layoutState.mLayoutDirection == LayoutState.LAYOUT_END条件代码为

if (layoutState.mLayoutDirection == LayoutState.LAYOUT_END) {return mSpans[(index++) % mSpanCount];
}

这样我们就可以在adapter里面根据position去区分是哪一行了

虽然上面代码能够有规律的显示数据了，但是当我们回滚时，会发现行数据会概率发现交换，通过打印代码发现，行数据发生交换的时候会调用getNextSpan函数我们修改的代码块，index++了，所以行数据会发生交换，既然知道了原因，那么我们来看看为什么会调用到这个函数吧

在fill(EPGRecyclerView.Recycler recycler, LayoutState layoutState,EPGRecyclerView.State state)这个函数里，有这样一段代码

final int spanIndex = mLazySpanLookup.getSpan(position);
StaggeredGridLayoutManager.Span currentSpan;
final boolean assignSpan = spanIndex == StaggeredGridLayoutManager.LayoutParams.INVALID_SPAN_ID;
if (assignSpan) {currentSpan = lp.mFullSpan ? mSpans[0] : getNextSpan(layoutState);mLazySpanLookup.setSpan(position, currentSpan);if (DEBUG) {Log.d(TAG, "assigned " + currentSpan.mIndex + " for " + position);}
}

当我们回滚时，按道理assignSpan应该是false的，spanIndex应该是有确切值的，而且spanIndex设置只有一处，这说明滚动时候在某个地方spanIndex被重置了

通过对代码的debug调试，原来是onLayoutChildren函数中有一个判断2个相邻的子View位置是否合理的逻辑，如果不合理的话会进行变量的重置，也就是上面的spanIndex

onLayoutChildren函数中有这样一段代码

boolean hasGaps = false;
if (shouldCheckForGaps && !state.isPreLayout() ) {//LogUtil.e("StaggeredGridLayoutManager", "onLayoutChildren", "state.isPreLayout() = false");final boolean needToCheckForGaps = mGapStrategy != GAP_HANDLING_NONE&& getChildCount() > 0&& (mLaidOutInvalidFullSpan || hasGapsToFix() != null);if (needToCheckForGaps) {removeCallbacks(mCheckForGapsRunnable);//LogUtil.e("StaggeredGridLayoutManager", "onLayoutChildren", "checkForGaps()3");if (checkForGaps()) {hasGaps = true;}}
}
if (state.isPreLayout()) {mAnchorInfo.reset();
}
mLastLayoutFromEnd = anchorInfo.mLayoutFromEnd;
mLastLayoutRTL = isLayoutRTL();
if (hasGaps) {//进行重置mAnchorInfo.reset();onLayoutChildren(recycler, state, false);
}

接下来我们来看看hasGapsToFix里面的一段代码

if (mShouldReverseLayout) {// ensure child's end is below nextChild's endint myEnd = mPrimaryOrientation.getDecoratedEnd(child);int nextEnd = mPrimaryOrientation.getDecoratedEnd(nextChild);if (myEnd < nextEnd) {return child; //i should have a better position} else if (myEnd == nextEnd) {compareSpans = true;}
} else {int myStart = mPrimaryOrientation.getDecoratedStart(child);int nextStart = mPrimaryOrientation.getDecoratedStart(nextChild);//如果当前View的起始处大于下一个View的起始处，会返回当前Viewif (myStart > nextStart) {return child; //i should have a better position} else if (myStart == nextStart) {compareSpans = true;}
}
if (compareSpans) {// equal, check span indices.StaggeredGridLayoutManager.LayoutParams nextLp = (StaggeredGridLayoutManager.LayoutParams) nextChild.getLayoutParams();if (lp.mSpan.mIndex - nextLp.mSpan.mIndex < 0 != preferredSpanDir < 0) {//preferredSpanDir < 0 为truereturn child;}
}

既然知道了原因，那么我们可以修改为不检测间隙，这样就避免了行数据交换的问题

①增加一个needCheckGap的标志位，置为false

//增加一个是否需要检测间隙的标志
private boolean needCheckGap = false;

②在hasGapsToFix函数起始处直接返回

View hasGapsToFix() {//不需要检测间隙 -zbjif (!needCheckGap) {return null;}...
}

③onLayoutChild函数去除检测间隙

boolean hasGaps = false;
//不检测间隙
if (shouldCheckForGaps && !state.isPreLayout() && hasGaps) {//LogUtil.e("StaggeredGridLayoutManager", "onLayoutChildren", "state.isPreLayout() = false");final boolean needToCheckForGaps = mGapStrategy != GAP_HANDLING_NONE&& getChildCount() > 0&& (mLaidOutInvalidFullSpan || hasGapsToFix() != null);if (needToCheckForGaps) {removeCallbacks(mCheckForGapsRunnable);//LogUtil.e("StaggeredGridLayoutManager", "onLayoutChildren", "checkForGaps()3");if (checkForGaps()) {hasGaps = true;}}
}

到这里，EPG功能布局已经差不多了，可以正常在手机上显示了，但是在安卓TV上，用遥控器左右键来进行选择时，会出现焦点不在同一行的问题，经过对安卓焦点规律的研究，发现是当滚动到屏幕后面的时候，后面已经没有实际的View了，这时候安卓系统会就近选择View，也就会出现左右切换焦点不在同一行的问题，解决这个问题也挺简单的，就是让RecyclerView额外多加载一层View，onLayoutChildren函数主要是进行子View的添加的，所以我们要从这里开始研究，该函数中有这样的代码块

// Layout end.
setLayoutStateDirection(LayoutState.LAYOUT_END);
testNum = 3;
fill(recycler, mLayoutState, state);
// Layout start.
setLayoutStateDirection(LayoutState.LAYOUT_START);
mLayoutState.mCurrentPosition = anchorInfo.mPosition + mLayoutState.mItemDirection;
fill(recycler, mLayoutState, state);

这里有2个fill函数，其实都是填充布局的，只不过第1个是为了做一些边缘检测处理，第二个才是真正的进行填充布局，接下来看看fill函数，StaggeredGridLayoutManager有一个判断某列或行是否还需要进行填充布局的标志mRemainingSpans，fiil函数填充布局的时候就会去对这个标志位做判断和赋值，以便用来加载子View

fill函数起始处会对mRemainingSpans做一个复位，也就是所有的列或行都需要进行添加子View

mRemainingSpans.set(0, mSpanCount, true);

接着在while循环中有这样一段代码，主要是用于检测该列或行末尾是否已经达到需要展示的位置，如果达到的话就将该列的标志位置为false，也就不再需要进行添加子View了

if (lp.mFullSpan) {updateAllRemainingSpans(mLayoutState.mLayoutDirection, targetLine);
} else {updateSpanNum = 1;updateRemainingSpans(currentSpan, mLayoutState.mLayoutDirection, targetLine);
}

private void updateRemainingSpans(StaggeredGridLayoutManager.Span span, int layoutDir, int targetLine) {final int deletedSize = span.getDeletedSize();final int line;if (layoutDir == LayoutState.LAYOUT_START) {line = span.getStartLine();if (line + deletedSize <= targetLine) {mRemainingSpans.set(span.mIndex, false);}} else {line = span.getEndLine();if (line - deletedSize >= targetLine) {mRemainingSpans.set(span.mIndex, false);}}
}

知道了填充布局的原理，那么我们就添加多一个BitSet标志位来使mRemainingSpans延迟置位，以达到多填充一层子View的目的

定义一个全局BitSet

/*** 增加一个延迟置位的标志，使其每列/行能添加多一个View */
private BitSet mDelayFlagSpans;

在setSpanCount(int spanCount)函数进行初始化

public void setSpanCount(int spanCount) {...mRemainingSpans = new BitSet(mSpanCount);mDelayFlagSpans = new BitSet(mSpanCount);... 
}

接着在fill函数起始处添加置位标志

mRemainingSpans.set(0, mSpanCount, true);
mDelayFlagSpans.set(0, mSpanCount, true);

在updateRemainingSpans函数中对mRemainingSpans做延迟操作

private void updateRemainingSpans(StaggeredGridLayoutManager.Span span, int layoutDir, int targetLine) {final int deletedSize = span.getDeletedSize();final int line;if (layoutDir == LayoutState.LAYOUT_START) {line = span.getStartLine();if (line + deletedSize <= targetLine) {if (mDelayFlagSpans.get(span.mIndex)) {mDelayFlagSpans.set(span.mIndex, false);} else {mRemainingSpans.set(span.mIndex, false);}}} else {line = span.getEndLine();if (line - deletedSize >= targetLine) {if (mDelayFlagSpans.get(span.mIndex)) {mDelayFlagSpans.set(span.mIndex, false);} else {mRemainingSpans.set(span.mIndex, false);}}}
}

到这里差不多了，我们可以运行了，但是通过打印发现当一滚动的时候position会一下子多加载了几十个，这样的结果我们是不能接受的，既然一滚动就会多加载，那么肯定是在滚动的时候会调用到fill函数，果不其然，在scrollBy函数中调用到了fill函数

int scrollBy(int dt, EPGRecyclerView.Recycler recycler, EPGRecyclerView.State state) {...int consumed = fill(recycler, mLayoutState, state);...
}

由于scrollBy函数会调用很多次，所以会额外加载了更多的子View，所以我们不能在fill函数里面对mDelayFlagSpans做复位操作，通过对代码的研究，我发现在3处地方加入mDelayFlagSpans的复位操作比较好，第1，2处是在onLayoutChild函数

// Layout end.
setLayoutStateDirection(LayoutState.LAYOUT_END);
mDelayFlagSpans.set(0, mSpanCount, true);
fill(recycler, mLayoutState, state);
// Layout start.
setLayoutStateDirection(LayoutState.LAYOUT_START);
mLayoutState.mCurrentPosition = anchorInfo.mPosition + mLayoutState.mItemDirection;
testNum = 4;
mDelayFlagSpans.set(0, mSpanCount, true);
fill(recycler, mLayoutState, state);

一开始布局的时候我们可以对mDelayFlagSpans进行复位，后面滚动的时候是不会在调用到此段代码，第3处是在RecyclerView的dispatchKeyEvent函数对按键做一下判断，并在StaggeredGridLayoutManager中加个对mDelayFlagSpans的复位函数以便在RecyclerView中调用

@Override
public boolean dispatchKeyEvent(KeyEvent event) {int keyCode = event.getKeyCode();// 这里只考虑水平移动的情况（垂直移动相同的解决方案）if (keyCode == KeyEvent.KEYCODE_DPAD_LEFT || keyCode == KeyEvent.KEYCODE_DPAD_RIGHT) {if (event.getAction() == KeyEvent.ACTION_DOWN) {if(getLayoutManager() instanceof StaggeredGridLayoutManager){StaggeredGridLayoutManager staggeredGridLayoutManager = (StaggeredGridLayoutManager)getLayoutManager();staggeredGridLayoutManager.resetDelayFlagSpans();}}}return super.dispatchKeyEvent(event);
}

StaggeredGridLayoutManager增加函数

public void resetDelayFlagSpans(){mDelayFlagSpans.set(0, mSpanCount, true);
}

到这里，我们已经基本完成了对StaggeredGridLayoutManager的修改，这里的修改只适合屏幕不做横竖屏切换的情况，因为去除了判断间隙，所以横竖屏切换会出现间隙