UI遍历中页面定义和动作事件筛选方法

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在Android UI遍历测试中，除传统的基于monkey的随机性测试外，基于模型的测试在测试覆盖率和可回溯性上表现更好，是目前热门的研究方向。在基于模型的测试中，对UI页面的定义和动作事件的筛选是十分重要而基础的工作。本文将介绍UI页面定义和动作事件筛选的具体方法。"}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"技术实现"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"本项目使用python。我们可以使用UI Automator来获取UI界面的层次树信息。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"View Tree"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"UI页面其实是一种树状结构的数据，称作view tree，其节点是每一个子view。每个子view一般包含'resource_id'、'scrollable'、'clickable'、'bounds'等信息，我们可以充分利用这些信息，来对页面进行定义以及动作事件的筛选。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/infoq\/d5\/d50e4297db5879f6ee2f93ab7fcca17d.png","alt":"图片","title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"为了便于操作，先将view tree树结构转成list列表形式，并保存子view的index索引等信息。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"text"},"content":[{"type":"text","text":"def get_view_list(view_tree):\n view_tree['parent'] = -1\n view_list = []\n view_tree_to_list(0, view_tree, view_list)\n self.last_acc_event['view_list'] = view_list\n return view_list\n\ndef view_tree_to_list(index, view_tree, view_list):\n tree_id = len(view_list)\n view_tree['temp_id'] = tree_id\n bounds = [[-1, -1], [-1, -1]]\n bounds[0][0] = view_tree['bounds'][0]\n bounds[0][1] = view_tree['bounds'][1]\n bounds[1][0] = view_tree['bounds'][2]\n bounds[1][1] = view_tree['bounds'][3]\n width = bounds[1][0] - bounds[0][0]\n height = bounds[1][1] - bounds[0][1]\n view_tree['size'] = \"%d*%d\" % (width, height)\n view_tree['index'] = index\n view_tree['bounds'] = bounds\n view_list.append(view_tree)\n children_ids = []\n for item in range(len(view_tree['children'])):\n child_tree = view_tree['children'][item]\n child_tree['parent'] = tree_id\n view_tree_to_list(item, child_tree, view_list)\n children_ids.append(child_tree['temp_id'])\n view_tree['children'] = children_ids"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"新页面的定义"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"由于App里有feed页等可无限刷新的页面，这就需要对页面进行定义区分，将类似的页面归为一类，避免UI遍历过程陷入无限循环的状态。这里，我们对页面有效信息提取成文本，并哈希成字符串作为该页面的唯一标识符。具体地，我们提取了每一个子view的'class'、'clickable'、'checked'、'scrollable'、'long-clickable'、'text'这些信息，将UI页面所有的子view信息组成文本，并用md5哈希成字符串。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"text"},"content":[{"type":"text","text":"def get_state_str(view_list):\n state_str_raw = get_state_str_raw(view_list)\n return md5(state_str_raw)\n\ndef get_state_str_raw(view_list):\n view_signatures = set()\n for view in view_list:\n view_signature = get_view_signature(view)\n if view_signature:\n view_signatures.add(view_signature)\n return \"%s{%s}\" % (self.foreground_activity, \",\".join(sorted(view_signatures)))\n\ndef get_view_signature(view_dict):\n view_text = view_dict['text']\n if view_text is None or len(view_text) > 50:\n view_text = \"None\"\n\n signature = \"[class]%s[text]%s[%s,%s,%s,%s]\" % \\\n (view_dict['class'],\n view_text,\n view_dict['clickable'],\n view_dict['checked'],\n view_dict['scrollable'],\n view_dict['long-clickable']\n )\n return signature\n"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"动作事件筛选"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我们遍历所有的子view，首先去掉'resource_id'为'android:id\/navigationBarBackground'、'android:id\/statusBarBackground'的导航栏的view，这在遍历测试中是不需要的。但有的app，它的导航栏view的'resource_id'不是这个，那就需要新加入过滤的内容，或者直接通过顶栏的座标过滤掉导航栏。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"过滤完系统的view事件之后，我们继续筛选，这里我们选取了'scrollable'或者'clickable'为true的子view。当然我们还可以筛选'enabled'、'focusable'等为true的子view事件，可以根据项目实际需要自行定义选择。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"text"},"content":[{"type":"text","text":"def get_possible_input(view_list):\n possible_events = []\n enabled_view_ids = []\n touch_exclude_view_ids = set()\n for view_dict in view_list:\n if view_dict['enabled'] and \\\n view_dict['resource_id'] not in \\\n ['android:id\/navigationBarBackground',\n 'android:id\/statusBarBackground']:\n enabled_view_ids.append(view_dict['temp_id'])\n\n for view_id in enabled_view_ids:\n if view_list[view_id]['scrollable']:\n possible_events.append(ScrollEvent(view=views_list[view_id], direction=\"UP\"))\n possible_events.append(ScrollEvent(view=views_list[view_id], direction=\"DOWN\"))\n possible_events.append(ScrollEvent(view=views_list[view_id], direction=\"LEFT\"))\n possible_events.append(ScrollEvent(view=views_list[view_id], direction=\"RIGHT\"))\n elif view_list[view_id]['clickable']:\n possible_events.append(TouchEvent(view=views_list[view_id]))\n touch_exclude_view_ids.add(view_id)\n # elif views_list[view_id]['enabled'] and \\\n # views_list[view_id]['focusable']:\n # possible_events.append(TouchEvent(view=views_list[view_id]))\n return possible_events"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我们如果想优先把'scrollable'的事件放在前面，可以分开筛选。有的时候向下的ScrollEvent事件是不必要的，我们也可以注释掉。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"text"},"content":[{"type":"text","text":"for view_id in enabled_view_ids:\n if views_list[view_id]['scrollable']:\n possible_events.append(ScrollEvent(view=views_list[view_id], direction=\"UP\"))\n # possible_events.append(ScrollEvent(view=views_list[view_id], direction=\"DOWN\"))\n possible_events.append(ScrollEvent(view=views_list[view_id], direction=\"LEFT\"))\n possible_events.append(ScrollEvent(view=views_list[view_id], direction=\"RIGHT\"))\n\nfor view_id in enabled_view_ids:\n if views_list[view_id]['clickable']:\n possible_events.append(TouchEvent(view=views_list[view_id]))\n touch_exclude_view_ids.add(view_id)"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"另外，我们还可以过滤一些不必要的子view事件。比如，有的子view的座标bounds超出了UI界面，这些界面外子view是不需要去测试遍历的。还有的子view的bounds会挤在一小块的像素内，这些也是不必要的。这里，我们设置如果bounds上下边界座标差小于5时，就过滤掉。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"text"},"content":[{"type":"text","text":"def filter_possible_input(possible_events,origin_dim=[1080, 1920]):\n filter_events = []\n for event in possible_events:\n # 过滤座标为负的值\n bounds = event.view[\"bounds\"]\n bounds = [bounds[0][0], bounds[0][1], bounds[1][0], bounds[1][1]]\n x_min = max(0, bounds[0])\n y_min = max(0, bounds[1])\n x_max = min(origin_dim[0], bounds[2])\n y_max = min(origin_dim[1], bounds[3])\n if x_min >= x_max or y_min >= y_max:\n continue\n # 更新bounds座标点\n event.view[\"bounds\"] = [[x_min,y_min],[x_max,y_max]]\n\n # 过滤小于5个像素的event\n if (y_max-y_min) < 5:\n pass\n else:\n filter_events.append(event)\n\n return filter_events"}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"总结"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"通过对页面的定义和动作事件的筛选，我们可以将不同的页面进行区分，相似的页面归为一类，筛选有效的event事件。在此基础上，我们可以构建图模型，将测试任务变为对有向图的遍历问题，在图模型上应用不同的算法，比如深度优先遍历、启发式搜索、深度学习或者强化学习算法等，对App进行充分的遍历测试。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"本文转载自：360技术（ID：qihoo_tech）"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"原文链接："},{"type":"link","attrs":{"href":"https:\/\/mp.weixin.qq.com\/s\/haY-9TBzn-g0g3lckQJ-MQ","title":"xxx","type":null},"content":[{"type":"text","text":"UI遍历中页面定义和动作事件筛选方法"}]}]}]}