TreeSaver 图片的定位

Treesaver 是浏览器大小尺寸敏感（size-sensitive）的，会就着当前的浏览器尺寸（browser size），选用不同的分栏表格（grid）做排版。不同排版效果下，图片出现的位置有啥规律，这就是本文要分析的内容：

一些典型的图片出现的规律：

首先我们看一些图片出现的规律：

一、当显示的区域只有两栏时，显示另外一个能贯穿两栏的图片，而不是之前贯穿三栏的图片。

这时候的图片相关CSS定义如下(使用Firefox的firebug监控到的解析后的html)：

<div data-sizes="double-16x9 double-4x3 double-1x1 double-3x4" 
class="container col-1 img w6col overflow double-16x9 zoomable" 
  style="bottom: auto;" data-figureindex="3">
  <div data-minwidth="518" data-sizes="double-16x9" data-minheight="363" class="">
        <img width="518" height="291" data-src="../img/10_395.jpg" src="../img/10_395.jpg">
        <p class="caption">
          (double)Jacob Wysocki in the title role, and John C. Reilly as Mr.
          Fitzgerald in Azazel Jacobs’ new film <em>Terri</em>.&nbsp; 
          <span class="credit">Art Takes Over5</span>
        </p>
</div></div>

二、三栏，图片在最上面，贯穿三栏：

这时候的图片相关CSS定义如下(使用Firefox的firebug监控到的解析后的html)：

<div data-sizes="triple-16x9 triple-4x3 triple-1x1 triple-3x4" 
  class="container col-1 img w9col overflow triple-16x9 zoomable" 
    style="bottom: auto;" data-figureindex="3">
   <div data-minwidth="788" data-sizes="triple-16x9" data-minheight="515" class="">
  <img width="788" height="443" data-src="../img/10_394.jpg" 
             src="../img/10_394.jpg">
  <p class="caption">
    (triple) Jacob Wysocki in the title role, and John C. Reilly as Mr.
    Fitzgerald in Azazel Jacobs’ new film <em>Terri</em>.&nbsp; 
                <span class="credit">Art Takes Over6</span>
  </p>
   </div>
</div>

三、四栏时图片出现在左下角，图片仍然是三栏

这时候的图片相关CSS定义如下(使用Firefox的firebug监控到的解析后的html)：

<div> 
    <div data-sizes="triple-16x9 triple-3x4 triple-1x1 triple-4x3" 
    class="container col-1 w9col img bottom triple-16x9 zoomable" style="top: auto;" 
    data-figureindex="3">
    <div data-minwidth="788" data-sizes="triple-16x9" data-minheight="515" class="">
  <img width="788" height="443" data-src="../img/10_394.jpg" src="../img/10_394.jpg">
  <p class="caption">
    (triple) Jacob Wysocki in the title role, and John C. Reilly as Mr.
    Fitzgerald in Azazel Jacobs’ new film <em>Terri</em>.&nbsp; 
    <span class="credit">Art Takes Over6</span>
  </p>
  </div>
  </div> 
    <div class="column col-1 flexed" style="min-height: 0px; bottom: 486px; height: 72px;">
      <h5 style="margin-top: 0px;">动感车身（前脸、保险杠、扰流板）</h5><p class="first">
  一些混排文字  
    </p></div> 
    <div class="column col-4 flexed" style="min-height: 0px; bottom: 486px; height: 72px;">
      <p class="first" style="margin-top: -54px;">
  一些混排文字  
  </p></div> 
    <div class="column col-7 flexed" style="min-height: 0px; bottom: 486px; height: 72px;">
    <p class="first" style="margin-top: -126px;">
  一些混排文字
  </p></div> 
</div>

四、三栏，刚好可以放下一个图

这时候的图片相关CSS定义如下(使用Firefox的firebug监控到的解析后的html)：

<div> 
    <div data-sizes="triple-16x9 triple-3x4 triple-1x1 triple-4x3" 
    class="container col-1 w9col img bottom triple-16x9 zoomable" style="top: auto;" data-figureindex="3">
    <div data-minwidth="788" data-sizes="triple-16x9" data-minheight="515" class="">
  <img width="788" height="443" data-src="../img/10_394.jpg" src="../img/10_394.jpg">
  <p class="caption">
    (triple) Jacob Wysocki in the title role, and John C. Reilly as Mr.
    Fitzgerald in Azazel Jacobs’ new film <em>Terri</em>.&nbsp; 
    <span class="credit">Art Takes Over6</span>
  </p>
</div></div> 
    <div class="column col-1 flexed" style="min-height: 0px; bottom: 486px;"> 
    </div> 
    <div class="column col-4 flexed" style="min-height: 0px; bottom: 486px;"> 
    </div> 
    <div class="column col-7 flexed" style="min-height: 0px; bottom: 486px;"> 
    </div> 
</div>
   

二、三这两个情况其实产生的html是一样的，只不过三产生的html中，填充图片空白位置的文字是空的，二产生的空白位置的文字是有的。

五、四栏图片，出现在最上面

这时候的图片相关CSS定义如下(使用Firefox的firebug监控到的解析后的html)：

<div data-sizes="quad-16x9 quad-4x3 quad-1x1 quad-3x4" 
       class="container col-1 img w12col overflow quad-16x9 zoomable" 
  style="bottom: auto;" data-figureindex="3">
    <div data-minwidth="1058" data-sizes="quad-16x9" data-minheight="667" class="">
        <img width="1058" height="595" data-src="../img/325_16813.jpg" src="../img/325_16813.jpg">
        <p class="caption">
          (quad) Jacob Wysocki in the title role, and John C. Reilly as Mr.
          Fitzgerald in Azazel Jacobs’ new film <em>Terri</em>.&nbsp; 
          <span class="credit">Art Takes Over7</span>
        </p>
</div></div>

在页面文件中我们可以看到下面的Html代码，显然我们可以在这里看到，同一个图片，我们预先定义了一些适合于不同设备的图片

<!--同一张图片 预先定义一些适用于不同设备的图片 -->
    <figure class="zoomable">
    
    <!-- 移动设备大小 -->
      <div data-minHeight="229" data-sizes="mobile-16x9"
        data-minWidth="280">
        <!-- 演示效果中没有这部分，就省略它 //-->
      </div>
      
    <!--小单栏 -->
      <div data-minHeight="160" data-sizes="small-single-16x9"
        data-minWidth="158">
      <!-- 演示效果中没有这部分，就省略它 //-->
      </div>
    <!-- 单栏 (single) -->
      <div data-minHeight="211" data-sizes="single-16x9"
        data-minWidth="248">
        <!-- 演示效果中没有这部分，就省略它 //-->
      </div>
     <!-- 小双栏 -->
      <div data-minHeight="262" data-sizes="small-double-16x9"
        data-minWidth="338">
        <!-- 演示效果中没有这部分，就省略它 //-->
      </div>
     <!-- 双栏(double) -->
      <div data-minHeight="363" data-sizes="double-16x9"
        data-minWidth="518">
        <img width="518"
          data-src="../img/10_395.jpg"
          height="291" />
        <p class="caption">
          (double)Jacob Wysocki in the title role, and John C. Reilly as Mr.
          Fitzgerald in Azazel Jacobs’ new film <em>Terri</em>.&nbsp; <span
            class="credit">Art Takes Over5</span>
        </p>
      </div>
   <!-- 19 显示器 16x9  三栏(triple)-->
<div data-minHeight="515" data-sizes="triple-16x9"
  data-minWidth="788">
  <img width="788"
    data-src="../img/10_394.jpg"
    height="443" />
  <p class="caption">
    (triple) Jacob Wysocki in the title role, and John C. Reilly as Mr.
    Fitzgerald in Azazel Jacobs’ new film <em>Terri</em>.&nbsp; <span
      class="credit">Art Takes Over6</span>
  </p>
</div>
      <!-- 四栏(quad) -->
      <div data-minHeight="667" data-sizes="quad-16x9" data-minWidth="1058">
        <img width="1058"
          data-src="../img/325_16813.jpg"
          height="595" />
        <p class="caption">
          (quad) Jacob Wysocki in the title role, and John C. Reilly as Mr.
          Fitzgerald in Azazel Jacobs’ new film <em>Terri</em>.&nbsp; <span
            class="credit">Art Takes Over7</span>
        </p>
      </div>
    <!-- fallback (备用)  -->
      <div class="img fallback keeptogether" data-sizes="fallback">
        <img  src="../img/test1.jpg"
          data-sizes="fallback" width="40" itemprop="thumb" height="40" />
        <p class="caption">
          (fallback) Jacob Wysocki in the title role, and John C. Reilly as Mr.
          Fitzgerald in Azazel Jacobs’ new film <em>Terri</em>.&nbsp; <span
            class="credit">Art Takes Over8</span>
        </p>
      </div>
    </figure>

同一个图片位置，我们上面看到预先定义了一些适合不同设备的不同大小的图片，每个图片的设置中描述了这个图片适合的大小，以及适合的分栏情况：

比如这个：<div data-minHeight="363" data-sizes="double-16x9"  data-minWidth="518">

就描述了适合于双栏：data-sizes="double-16x9"，最小高度要求：data-minHeight="363"，最小宽度要求：data-minWidth="518"

TreeSaver 在运算出适合用几栏来显示后， 根据这里定义的适合的栏数，以及最小长宽，选择合适的图来显示。

图片的位置从上面分析来看，只有两种：

overflow 和 bottom 。 对应的样式：  style="top: auto;"  和 style="bottom: auto;"

TreeSaver的源代码中有下面代码：

/**
 * @param {!Element} container
 * @param {!treesaver.layout.Figure} figure
 * @param {!treesaver.layout.Grid.ContainerMap} map
 * @param {?number} lineHeight
 * @return {boolean} True if the figure fit within the container.
 */
treesaver.layout.Page.fillContainer = function(container, figure, map,
    lineHeight) {
  var size, figureSize,
      containerHeight, sibling,
      maxContainerHeight,
      anchoredTop = true;
  size = map.size;
  figureSize = map.figureSize;
  if (goog.DEBUG) {
    if (!size) {
      treesaver.debug.error('Empty size!');
    }
    if (!figureSize) {
      treesaver.debug.error('Empty figureSize!');
    }
  }
  maxContainerHeight = treesaver.dimensions.getOffsetHeight(container);
  // Do any content switching that needs to happen
  figureSize.applySize(container, size);
  // If the container is fixed, then we are done no matter what
  if (!map.flexible) {
    return true;
  }
  // Adjust flexible containers
  // Unhinge from a side before measuring
  if (treesaver.dom.hasClass(container, 'bottom')) {
    anchoredTop = false;
    container.style.top = 'auto';
  }
  else {
    container.style.bottom = 'auto';
  }
  containerHeight = treesaver.dimensions.getOffsetHeight(container);
  // Did not fit :(
  // TODO: Use something better than parent height
  if (containerHeight > maxContainerHeight) {
    treesaver.debug.info('Container failure: ' + containerHeight + ':' + maxContainerHeight);
    if (goog.DEBUG) {
      container.setAttribute('data-containerHeight', containerHeight);
      container.setAttribute('data-maxHeight', maxContainerHeight);
      container.setAttribute('data-attemptedSize', size);
    }
    // Revert after failure
    figureSize.revertSize(container, size);
    // TODO: Return style.bottom & style.top to originals?
    return false;
  }
  // Round to nearest for column adjustment to maintain grid
  if (lineHeight && containerHeight % lineHeight) {
    containerHeight = treesaver.dimensions.roundUp(containerHeight, lineHeight);
  }
  // Go through this containers siblings, adjusting their sizes
  sibling = container;
  while ((sibling = sibling.nextSibling)) {
    if (sibling.nodeType !== 1) {
      // Ignore non-elements
      continue;
    }
    // Cast for compiler
    sibling = /** @type {!Element} */ (sibling);
    // Don't touch fixed items
    if (treesaver.dom.hasClass(sibling, 'fixed')) {
      continue;
    }
    if (treesaver.dom.hasClass(sibling, 'column') ||
        treesaver.dom.hasClass(sibling, 'container') ||
        treesaver.dom.hasClass(sibling, 'group')) {
      // Add a flag for debugging / later detection
      treesaver.dom.addClass(sibling, 'flexed');
      // Make sure we don't go negative
      if (treesaver.dimensions.getOffsetHeight(sibling) <= containerHeight) {
        treesaver.debug.info('Sibling shrunk to zero height: ' + sibling);
        // TODO: Remove from tree?
        treesaver.dimensions.setCssPx(sibling, 'height', 0);
      }
      else {
        // Since items are always absolutely positioned, we can
        // adjust the position of the column directly based on it's
        // offsets
        if (anchoredTop) {
          treesaver.dimensions.setCssPx(sibling, 'top',
            treesaver.dimensions.getOffsetTop(sibling) + containerHeight);
        }
        else {
          // Compute the current 'bottom' value by using the parent's offsetHeight
          treesaver.dimensions.setCssPx(sibling, 'bottom',
            treesaver.dimensions.getOffsetHeight(sibling.offsetParent) -
            (treesaver.dimensions.getOffsetTop(sibling) + treesaver.dimensions.getOffsetHeight(sibling)) + containerHeight);
        }
      }
    }
  }
  return true;
};

注意其中的这一段：

  if (treesaver.dom.hasClass(container, 'bottom')) {
    anchoredTop = false;
    container.style.top = 'auto';
  }
  else {
    container.style.bottom = 'auto';
  }

这意味着，默认情况下，我们是把图片显示在上面。 只有在 treesaver.dom.hasClass(container, 'bottom') 时才会显示图片在左下角。
treesaver.dom.hasClass 函数如下：

/**
 * Check if an element has the given class
 * Hat Tip: Dean Edwards http://dean.edwards.name/IE7/caveats/
 *
 * @param {!Element|!HTMLDocument} el
 * @param {!string} className
 * @return {boolean} True if the element has that class.
 */
treesaver.dom.hasClass = function(el, className) {
  var regexp = new RegExp('(^|\\s)' + className + '(\\s|$)');
  return !!(el.className && regexp.test(el.className));
};

即，我们通过判断元素的class 中是否存在 'bottom' 这个样式定义来决定是否该如何显示的。

我们对上面代码加断点，并监控 container.className 的值，会看到它的值是：类似如下的："container col-7 w6col img double-4x3"

我们在资源文件中搜索container col-7 w6col img，会看到两个Grid中有如下定义：

<!-- 非第一页的奇数页 //-->
  <div class="grid w12col feature1 odd no-page-1"> 
   <div class="runningtitle"> 
    <span> 
     Nomad Editions
    </span> 
    <span data-bind="vertical"> 
     vertical
    </span> 
   </div> 
   <div> 
    <div class="container col-1 w3col pullquote bottom" data-sizes="pullquote"> 
    </div> 
    <div class="column col-1"> 
    </div> 
   </div> 
   <div class="column col-4"> 
   </div> 
   <div> 
    <div class="container col-7 w6col img" data-sizes="double-16x9 double-3x4 double-1x1 double-4x3"> 
    </div> 
    <div class="column col-7 bottom"> 
    </div> 
    <div class="column col-10 bottom"> 
    </div> 
   </div> 
  </div> 
<!-- 非第一页的偶数页 //-->
<div class="grid w12col feature2 even no-page-1"> 
   <div class="runningtitle"> 
    <span> 
     Nomad Editions
    </span> 
    <span data-bind="vertical"> 
     vertical
    </span> 
   </div> 
   <div> 
    <div class="container col-1 w3col pullquote bottom" data-sizes="pullquote"> 
    </div> 
    <div class="column col-1"> 
    </div> 
   </div> 
   <div class="column col-4"> 
   </div> 
   <div> 
    <div class="container col-7 w6col img" data-sizes="double-16x9 double-3x4 double-1x1 double-4x3"> 
    </div> 
    <div class="column col-7"> 
    </div> 
    <div class="column col-10"> 
    </div> 
   </div> 
  </div> 

从这里我的猜测整个处理流程是

根据Grid的适用场景，我们选择合适的Grid，

在Grid内部，我们会根据我们的配置来显示这个Grid。

所以理解Grid内部工作原理，如何选择Grid会是理解TreeSaver的一个关键。