OverView
GridLayout是Android4.0引入的网格控件，可以方便地实现网格式布局，减少嵌套层级，这周看下GridLayout具体的工作原理。
onMeasure
照例从测量阶段开始，看下GridLayout在测量阶段进行了什么操作：
@Override
protected void onMeasure(int widthSpec, int heightSpec) {
consistencyCheck();
/** If we have been called by {@link View#measure(int, int)}, one of width or height
* is likely to have changed. We must invalidate if so. */
invalidateValues();
int hPadding = getPaddingLeft() + getPaddingRight();
int vPadding = getPaddingTop() + getPaddingBottom();
int widthSpecSansPadding = adjust( widthSpec, -hPadding);
int heightSpecSansPadding = adjust(heightSpec, -vPadding);
measureChildrenWithMargins(widthSpecSansPadding, heightSpecSansPadding, true);
int widthSansPadding;
int heightSansPadding;
// Use the orientation property to decide which axis should be laid out first.
if (mOrientation == HORIZONTAL) {
widthSansPadding = mHorizontalAxis.getMeasure(widthSpecSansPadding);
measureChildrenWithMargins(widthSpecSansPadding, heightSpecSansPadding, false);
heightSansPadding = mVerticalAxis.getMeasure(heightSpecSansPadding);
} else {
heightSansPadding = mVerticalAxis.getMeasure(heightSpecSansPadding);
measureChildrenWithMargins(widthSpecSansPadding, heightSpecSansPadding, false);
widthSansPadding = mHorizontalAxis.getMeasure(widthSpecSansPadding);
}
int measuredWidth = Math.max(widthSansPadding + hPadding, getSuggestedMinimumWidth());
int measuredHeight = Math.max(heightSansPadding + vPadding, getSuggestedMinimumHeight());
setMeasuredDimension(
resolveSizeAndState(measuredWidth, widthSpec, 0),
resolveSizeAndState(measuredHeight, heightSpec, 0));
}
GridLayout的onMeasure()方法并不长，可以分成以下3个部分：
- 兼容性检查、参数校验、padding的计算与调整
- 对子View进行第一次测量
- 对子View进行第二次测量
其中对子View的测量是通过调用measureChildrenWithMargins()实现的，看下这个函数的实现：
private void measureChildrenWithMargins(int widthSpec, int heightSpec, boolean firstPass) {
for (int i = 0, N = getChildCount(); i < N; i++) {
View c = getChildAt(i);
if (c.getVisibility() == View.GONE) continue;
LayoutParams lp = getLayoutParams(c);
if (firstPass) {
measureChildWithMargins2(c, widthSpec, heightSpec, lp.width, lp.height);
} else {
boolean horizontal = (mOrientation == HORIZONTAL);
Spec spec = horizontal ? lp.columnSpec : lp.rowSpec;
if (spec.getAbsoluteAlignment(horizontal) == FILL) {
Interval span = spec.span;
Axis axis = horizontal ? mHorizontalAxis : mVerticalAxis;
int[] locations = axis.getLocations();
int cellSize = locations[span.max] - locations[span.min];
int viewSize = cellSize - getTotalMargin(c, horizontal);
if (horizontal) {
measureChildWithMargins2(c, widthSpec, heightSpec, viewSize, lp.height);
} else {
measureChildWithMargins2(c, widthSpec, heightSpec, lp.width, viewSize);
}
}
}
}
}
在这个函数中，可以看到有1个for循环来遍历子View，然后通过firstPass参数判断是第一轮测量还是第二轮测量，如果是第一轮测量，传入的参数是lp.width/lp.height来对子View进行测量；
关键在于第二轮测量的处理，可以看到如果spec.getAbsoluteAlignment(horizontal) == FILL这个条件不成立的话，第二轮测量实际上是没有对子View进行测量操作的。
我们来看下spec.getAbsoluteAlignment(horizontal) 这个的实现：
private Alignment getAbsoluteAlignment(boolean horizontal) {
if (alignment != UNDEFINED_ALIGNMENT) {
return alignment;
}
if (weight == 0f) {
return horizontal ? START : BASELINE;
}
return FILL;
}
可以看到，这个函数的返回结果和alignment与weight的值有关，weight的值就是我们设置给某个单元格的行/列权重，alignment与设置的gravity有关：
static Alignment getAlignment(int gravity, boolean horizontal) {
int mask = horizontal ? HORIZONTAL_GRAVITY_MASK : VERTICAL_GRAVITY_MASK;
int shift = horizontal ? AXIS_X_SHIFT : AXIS_Y_SHIFT;
int flags = (gravity & mask) >> shift;
switch (flags) {
case (AXIS_SPECIFIED | AXIS_PULL_BEFORE):
return horizontal ? LEFT : TOP;
case (AXIS_SPECIFIED | AXIS_PULL_AFTER):
return horizontal ? RIGHT : BOTTOM;
case (AXIS_SPECIFIED | AXIS_PULL_BEFORE | AXIS_PULL_AFTER):
return FILL;
case AXIS_SPECIFIED:
return CENTER;
case (AXIS_SPECIFIED | AXIS_PULL_BEFORE | RELATIVE_LAYOUT_DIRECTION):
return START;
case (AXIS_SPECIFIED | AXIS_PULL_AFTER | RELATIVE_LAYOUT_DIRECTION):
return END;
default:
return UNDEFINED_ALIGNMENT;
}
}
由此可以得出，第二轮测量，实际上是根据gravity和weight值的设定，将多余的空间再次分配给子单元格
onLayout()
@Override
protected void onLayout(boolean changed, int left, int top, int right, int bottom) {
consistencyCheck();
int targetWidth = right - left;
int targetHeight = bottom - top;
int paddingLeft = getPaddingLeft();
int paddingTop = getPaddingTop();
int paddingRight = getPaddingRight();
int paddingBottom = getPaddingBottom();
mHorizontalAxis.layout(targetWidth - paddingLeft - paddingRight);
mVerticalAxis.layout(targetHeight - paddingTop - paddingBottom);
int[] hLocations = mHorizontalAxis.getLocations();
int[] vLocations = mVerticalAxis.getLocations();
for (int i = 0, N = getChildCount(); i < N; i++) {
View c = getChildAt(i);
if (c.getVisibility() == View.GONE) continue;
LayoutParams lp = getLayoutParams(c);
Spec columnSpec = lp.columnSpec;
Spec rowSpec = lp.rowSpec;
Interval colSpan = columnSpec.span;
Interval rowSpan = rowSpec.span;
int x1 = hLocations[colSpan.min];
int y1 = vLocations[rowSpan.min];
int x2 = hLocations[colSpan.max];
int y2 = vLocations[rowSpan.max];
int cellWidth = x2 - x1;
int cellHeight = y2 - y1;
int pWidth = getMeasurement(c, true);
int pHeight = getMeasurement(c, false);
Alignment hAlign = columnSpec.getAbsoluteAlignment(true);
Alignment vAlign = rowSpec.getAbsoluteAlignment(false);
Bounds boundsX = mHorizontalAxis.getGroupBounds().getValue(i);
Bounds boundsY = mVerticalAxis.getGroupBounds().getValue(i);
// Gravity offsets: the location of the alignment group relative to its cell group.
int gravityOffsetX = hAlign.getGravityOffset(c, cellWidth - boundsX.size(true));
int gravityOffsetY = vAlign.getGravityOffset(c, cellHeight - boundsY.size(true));
int leftMargin = getMargin(c, true, true);
int topMargin = getMargin(c, false, true);
int rightMargin = getMargin(c, true, false);
int bottomMargin = getMargin(c, false, false);
int sumMarginsX = leftMargin + rightMargin;
int sumMarginsY = topMargin + bottomMargin;
// Alignment offsets: the location of the view relative to its alignment group.
int alignmentOffsetX = boundsX.getOffset(this, c, hAlign, pWidth + sumMarginsX, true);
int alignmentOffsetY = boundsY.getOffset(this, c, vAlign, pHeight + sumMarginsY, false);
int width = hAlign.getSizeInCell(c, pWidth, cellWidth - sumMarginsX);
int height = vAlign.getSizeInCell(c, pHeight, cellHeight - sumMarginsY);
int dx = x1 + gravityOffsetX + alignmentOffsetX;
int cx = !isLayoutRtl() ? paddingLeft + leftMargin + dx :
targetWidth - width - paddingRight - rightMargin - dx;
int cy = paddingTop + y1 + gravityOffsetY + alignmentOffsetY + topMargin;
if (width != c.getMeasuredWidth() || height != c.getMeasuredHeight()) {
c.measure(makeMeasureSpec(width, EXACTLY), makeMeasureSpec(height, EXACTLY));
}
c.layout(cx, cy, cx + width, cy + height);
}
}
GridLayout的onLayout()函数乍看起来有点长，但明确onLayout()阶段目的之后就可以很好地梳理该函数的实现，onLayout()阶段的目的是确定好各个子View的位置，而对于GridLayout来说，子View的位置是通过x行y列这样的方式来设置的，因此只要计算出x行y列相应的坐标即可。
在onLayout()函数中需要特别注意的一点是，如果某个子View算出来的width和测量得到的width不一致时，会将算出的width传入子View再调用1次measure
onDraw()
GridLayout作为容器布局，也没有重写onDraw()函数。
总结
GridLayout的实现相对简单，主要注意几个点：
- 测量阶段会循环遍历2次子View，但第2次循环不一定会对子View进行测量
- 布局阶段有可能还会对子View进行measure()方法的调用。
