目前比较有名的Uitest框架有Uiautomator/Robotium/Appium，由于一直对webview元素的获取和操作比较好奇，另外Robotium代码量也不是很大，因此打算学习一下。
一.环境准备以及初始化
用来说明的用例采用的是Robotium官网的一个tutorial用例-Notepad
@RunWith(AndroidJUnit4.class)
public class NotePadTest { private static final String NOTE_1 = "Note 1";
private static final String NOTE_2 = "Note 2"; @Rule
public ActivityTestRule<NotesList> activityTestRule =
new ActivityTestRule<>(NotesList.class); private Solo solo; @Before
public void setUp() throws Exception {
//setUp() is run before a test case is started.
//This is where the solo object is created.
solo = new Solo(InstrumentationRegistry.getInstrumentation(),
activityTestRule.getActivity());
} @After
public void tearDown() throws Exception {
//tearDown() is run after a test case has finished.
//finishOpenedActivities() will finish all the activities that have been opened during the test execution.
solo.finishOpenedActivities();
} @Test
public void testAddNote() throws Exception {
//Unlock the lock screen
solo.unlockScreen();
//Click on action menu item add
solo.clickOnView(solo.getView(R.id.menu_add));
//Assert that NoteEditor activity is opened
solo.assertCurrentActivity("Expected NoteEditor Activity", NoteEditor.class);
//In text field 0, enter Note 1
solo.enterText(0, NOTE_1);
//Click on action menu item Save
solo.clickOnView(solo.getView(R.id.menu_save));
//Click on action menu item Add
solo.clickOnView(solo.getView(R.id.menu_add));
//In text field 0, type Note 2
solo.typeText(0, NOTE_2);
//Click on action menu item Save
solo.clickOnView(solo.getView(R.id.menu_save));
//Takes a screenshot and saves it in "/sdcard/Robotium-Screenshots/".
solo.takeScreenshot();
//Search for Note 1 and Note 2
boolean notesFound = solo.searchText(NOTE_1) && solo.searchText(NOTE_2);
//To clean up after the test case
deleteNotes();
//Assert that Note 1 & Note 2 are found
assertTrue("Note 1 and/or Note 2 are not found", notesFound);
} }
在进行初始化时，Solo对象依赖Instrumentation对象以及被测应用的Activity对象，在这里是NotesList，然后所有的UI操作都依赖这个Solo对象。
二.Native控件解析与操作
1.Native控件解析
看一个标准的操作：solo.clickOnView(solo.getView(R.id.menu_save));
solo点击id为menu_save的控件，其中clickOnView传入参数肯定为menu_save的view对象，那这个是怎么获取的呢？
由于调用比较深，因此直接展示关键方法
public View waitForView(int id, int index, int timeout, boolean scroll) {
HashSet uniqueViewsMatchingId = new HashSet();
long endTime = SystemClock.uptimeMillis() + (long)timeout;
while(SystemClock.uptimeMillis() <= endTime) {
this.sleeper.sleep();
Iterator i$ = this.viewFetcher.getAllViews(false).iterator();
while(i$.hasNext()) {
View view = (View)i$.next();
Integer idOfView = Integer.valueOf(view.getId());
if(idOfView.equals(Integer.valueOf(id))) {
uniqueViewsMatchingId.add(view);
if(uniqueViewsMatchingId.size() > index) {
return view;
}
}
}
if(scroll) {
this.scroller.scrollDown();
}
}
return null;
}
这个方法是先去获取所有的View： this.viewFetcher.getAllViews(false)，然后通过匹配id来获取正确的View。
那Robotium是怎么获取到所有的View呢？这就要看看viewFetcher里的实现了。
public ArrayList<View> getAllViews(boolean onlySufficientlyVisible) {
View[] views = this.getWindowDecorViews();
ArrayList allViews = new ArrayList();
View[] nonDecorViews = this.getNonDecorViews(views);
View view = null;
if(nonDecorViews != null) {
for(int ignored = 0; ignored < nonDecorViews.length; ++ignored) {
view = nonDecorViews[ignored];
try {
this.addChildren(allViews, (ViewGroup)view, onlySufficientlyVisible);
} catch (Exception var9) {
;
}
if(view != null) {
allViews.add(view);
}
}
}
if(views != null && views.length > 0) {
view = this.getRecentDecorView(views);
try {
this.addChildren(allViews, (ViewGroup)view, onlySufficientlyVisible);
} catch (Exception var8) {
;
}
if(view != null) {
allViews.add(view);
}
}
return allViews;
}
需要说明的是，DecorView是整个ViewTree的最顶层View，它是一个FrameLayout布局，代表了整个应用的界面。
从上面的代码可以看到，allViews包括WindowDecorViews，nonDecorViews，RecentDecorView。所以，我对这三个封装比较感兴趣，他们是怎么拿到WindowDecorViews，nonDecorViews，RecentDecorView的呢？
继续看代码，可以看到以下方法（看注释）
// 获取 DecorViews
public View[] getWindowDecorViews() {
try {
Field viewsField = windowManager.getDeclaredField("mViews");
Field instanceField = windowManager.getDeclaredField(this.windowManagerString);
viewsField.setAccessible(true);
instanceField.setAccessible(true);
Object e = instanceField.get((Object)null);
View[] result;
if(VERSION.SDK_INT >= 19) {
result = (View[])((ArrayList)viewsField.get(e)).toArray(new View[0]);
} else {
result = (View[])((View[])viewsField.get(e));
} return result;
} catch (Exception var5) {
var5.printStackTrace();
return null;
}
} // 获取NonDecorViews
private final View[] getNonDecorViews(View[] views) {
View[] decorViews = null;
if(views != null) {
decorViews = new View[views.length];
int i = 0; for(int j = 0; j < views.length; ++j) {
View view = views[j];
if(!this.isDecorView(view)) {
decorViews[i] = view;
++i;
}
}
} return decorViews;
} // 获取RecentDecorView
public final View getRecentDecorView(View[] views) {
if(views == null) {
return null;
} else {
View[] decorViews = new View[views.length];
int i = 0; for(int j = 0; j < views.length; ++j) {
View view = views[j];
if(this.isDecorView(view)) {
decorViews[i] = view;
++i;
}
} return this.getRecentContainer(decorViews);
}
}
其中DecorViews就不用多说了，通过反射拿到一个里面的元素都是View的List，而NonDecorViews则是通过便利DectorViews进行过滤，nameOfClass 不满足要求的，则为NonDecorViews
String nameOfClass = view.getClass().getName();
return nameOfClass.equals("com.android.internal.policy.impl.PhoneWindow$DecorView") || nameOfClass.equals("com.android.internal.policy.impl.MultiPhoneWindow$MultiPhoneDecorView") || nameOfClass.equals("com.android.internal.policy.PhoneWindow$DecorView");
而recentlyView则通过以下条件进行判断，满足则为recentlyView
view != null && view.isShown() && view.hasWindowFocus() && view.getDrawingTime() > drawingTime
2.Native控件解析
依旧说的是这个操作：solo.clickOnView(solo.getView(R.id.menu_save));接下来要看的是clickOnView的封装了。
这部分实现相对简单很多了，获取控件坐标的*X,Y值后，利用instrumentation的sendPointerSync来完成点击/长按操作
public void clickOnScreen(float x, float y, View view) {
boolean successfull = false;
int retry = 0;
SecurityException ex = null;
while(!successfull && retry < 20) {
long downTime = SystemClock.uptimeMillis();
long eventTime = SystemClock.uptimeMillis();
MotionEvent event = MotionEvent.obtain(downTime, eventTime, 0, x, y, 0);
MotionEvent event2 = MotionEvent.obtain(downTime, eventTime, 1, x, y, 0);
try {
this.inst.sendPointerSync(event);
this.inst.sendPointerSync(event2);
successfull = true;
} catch (SecurityException var16) {
ex = var16;
this.dialogUtils.hideSoftKeyboard((EditText)null, false, true);
this.sleeper.sleep(300);
++retry;
View identicalView = this.viewFetcher.getIdenticalView(view);
if(identicalView != null) {
float[] xyToClick = this.getClickCoordinates(identicalView);
x = xyToClick[0];
y = xyToClick[1];
}
}
}
if(!successfull) {
Assert.fail("Click at (" + x + ", " + y + ") can not be completed! (" + (ex != null?ex.getClass().getName() + ": " + ex.getMessage():"null") + ")");
}
}
3.总结：
从源码中可以看出，其实native控件操作的思想是这样的。
通过android.view.windowManager获取到所有的view，然后经过过滤得到自己需要的view，最后通过计算view的 Coordinates得到*坐标，最后依赖instrument来完成操作。
三.Webview的解析与操作
webview的解析需要利用JS注入获取到Web页面的元素，进行过滤后再进行操作。
webview的调试环境可以利用inspect来进行，具体参考此文章：http://www.cnblogs.com/sunshq/p/4111610.html
在这里进行说明的解析操作代码为：
ArrayList<WebElement> webElements = solo.getCurrentWebElements(By.className("ns-video ns-icon"));
solo.clickOnWebElement(webElements.get(0));
这段代码很好理解，取出className为"btn btn-block primary"的元素，并点击第一个。在这里，元素的可操作对象为WebElement.
debug界面为：
在具体debug代码前，我们有必要先了解一下解析Webview的流程应该是怎样的（尽管我是看代码了解的，但先把流程说一下方便解说），不然很可能会云里雾里。流程如下：
1. 获取所有的view，过滤出webview
2.初始化javascript环境
3.加载本地js并注入
4.WebElment操作
接下来，自然而然，带着目的去看代码，就可以很清楚每一步在做什么了。
1. 获取所有的view，过滤出webview
（1）直接跳到关键代码，首先要看的是By是用来干嘛的。通过查看源码，可以发现，其实By是一个Java bean，里面封装了Id/CssSelector/ClassName/Text等等属性，可以理解为WebView中的目标对象封装。
public boolean executeJavaScript(By by, boolean shouldClick) {
return by instanceof Id?this.executeJavaScriptFunction("id(\"" + by.getValue() + "\", \"" + shouldClick + "\");"):(by instanceof Xpath?this.executeJavaScriptFunction("xpath(\"" + by.getValue() + "\", \"" + shouldClick + "\");"):(by instanceof CssSelector?this.executeJavaScriptFunction("cssSelector(\"" + by.getValue() + "\", \"" + shouldClick + "\");"):(by instanceof Name?this.executeJavaScriptFunction("name(\"" + by.getValue() + "\", \"" + shouldClick + "\");"):(by instanceof ClassName?this.executeJavaScriptFunction("className(\"" + by.getValue() + "\", \"" + shouldClick + "\");"):(by instanceof Text?this.executeJavaScriptFunction("textContent(\"" + by.getValue() + "\", \"" + shouldClick + "\");"):(by instanceof TagName?this.executeJavaScriptFunction("tagName(\"" + by.getValue() + "\", \"" + shouldClick + "\");"):false))))));
}
private boolean executeJavaScriptFunction(final String function) {
ArrayList webViews = this.viewFetcher.getCurrentViews(WebView.class, true);
final WebView webView = (WebView)this.viewFetcher.getFreshestView((ArrayList)webViews);
if(webView == null) {
return false;
} else {
final String javaScript = this.setWebFrame(this.prepareForStartOfJavascriptExecution(webViews));
this.inst.runOnMainSync(new Runnable() {
public void run() {
if(webView != null) {
webView.loadUrl("javascript:" + javaScript + function);
}
}
});
return true;
}
}
executeJavaScript获取到的是对应的执行方法调用语句，这个根据自己定位的元素决定，在这，我的为："className(\"ns-video ns-icon\", \"false\");"
（2）getCurrentViews，获取所有当前View
public <T extends View> ArrayList<T> getCurrentViews(Class<T> classToFilterBy, boolean includeSubclasses, View parent) {
ArrayList filteredViews = new ArrayList();
ArrayList allViews = this.getViews(parent, true);
Iterator i$ = allViews.iterator();
while(true) {
View view;
Class classOfView;
do {
do {
if(!i$.hasNext()) {
allViews = null;
return filteredViews;
}
view = (View)i$.next();
} while(view == null);
classOfView = view.getClass();
} while((!includeSubclasses || !classToFilterBy.isAssignableFrom(classOfView)) && (includeSubclasses || classToFilterBy != classOfView));
filteredViews.add(classToFilterBy.cast(view));
}
}
其中classToFilterBy为android.webkit.Webview这个类，所作的操作为调用获取所有View（跟navitive调用的方法一致），包括控件view跟webView，如图所示
然后逐一过滤出，条件为(!includeSubclasses || !classToFilterBy.isAssignableFrom(classOfView)) && (includeSubclasses || classToFilterBy != classOfView))。在这里加个题外话，因为robotium默认的是android.webkit.Webview，因此如果你用robotium去解析操作第三方内核的Webview，是会失败的，解决办法就是改源码？
2.初始化javascript环境
看（1）的代码：this.setWebFrame(this.prepareForStartOfJavascriptExecution(webViews));
在这里会初始化一个robotiumWebCLient, 并将webChromeClient设置成RobotiumWebClient.this.robotiumWebClient，由于我对这一块也不熟悉，不太清楚这一块的目的，因此不赘述，姑且认为是执行js注入的环境初始化。
3.加载js脚本并注入
private String getJavaScriptAsString() {
InputStream fis = this.getClass().getResourceAsStream("RobotiumWeb.js");
StringBuffer javaScript = new StringBuffer();
try {
BufferedReader e = new BufferedReader(new InputStreamReader(fis));
String line = null;
while((line = e.readLine()) != null) {
javaScript.append(line);
javaScript.append("\n");
}
e.close();
return javaScript.toString();
} catch (IOException var5) {
throw new RuntimeException(var5);
}
}
这一块就没多少要说的了，就是把本地的js脚本加载进来，方便执行，最后在异步线程中将js注入，参考（1）中的webView.loadUrl("javascript:" + javaScript + function);
在这里可以展示一下我这边注入的js是怎样的(只展示结构，具体方法内容略掉)：
javascript:/**
* Used by the web methods.
*
* @author Renas Reda, renas.reda@robotium.com
*
*/ function allWebElements() {
...
} function allTexts() {
...
} function clickElement(element){
var e = document.createEvent('MouseEvents');
e.initMouseEvent('click', true, true, window, 1, 0, 0, 0, 0, false, false, false, false, 0, null);
element.dispatchEvent(e);
} function id(id, click) {
...
} function xpath(xpath, click) {
...
} function cssSelector(cssSelector, click) {
...
} function name(name, click) {
...
} function className(nameOfClass, click) {
...
} function textContent(text, click) {
...
} function tagName(tagName, click) {
...
} function enterTextById(id, text) {
...
} function enterTextByXpath(xpath, text) {
...
} function enterTextByCssSelector(cssSelector, text) {
...
} function enterTextByName(name, text) {
...
} function enterTextByClassName(name, text) {
...
} function enterTextByTextContent(textContent, text) {
...
} function enterTextByTagName(tagName, text) {
...
} function promptElement(element) {
...
} function promptText(element, range) {
...
} function finished(){
prompt('robotium-finished');
}
className("ns-video ns-icon", "false");
4.WebElment操作
接下来便是元素操作了，在这里的操作对象是WebElment，获取到下X，Y坐标进行对应操作就可以了。
总结：
这篇文章展示了robotium是如何去识别控件跟webview元素的，这个基本上是一个框架能用与否的关键，有兴趣造*或者想学习robotium源码的可以多多参考。