Android DeepLink原理与应用（2）
3.Android框架层如何实现DeepLink？
上一篇提到Android使用DeepLink跳转有两种方式：Intent跳转、WebView网页跳转。这篇研究一下Android框架如何实现DeepLink。以下Android源代码分析版本为Android7.1。
（1）Intent跳转
这其实是一个Intent filter + start Activity的过程，使用的Intent Catagory是Intent.CATEGORY_BROWSABLE，从这入手。先在Android框架层代码中搜索一下看看哪些地方处理了Intent.CATEGORY_BROWSABLE？发现在处理Intent过滤的工具类中有处理：
frameworks/base/core/java/android/content/IntentFilter.java：
/**
* Return if this filter handles HTTP or HTTPS data URIs.
*
* @return True if the filter handles ACTION_VIEW/CATEGORY_BROWSABLE,
* has at least one HTTP or HTTPS data URI pattern defined, and optionally
* does not define any non-http/https data URI patterns.
*
* This will check if if the Intent action is {@link android.content.Intent#ACTION_VIEW} and
* the Intent category is {@link android.content.Intent#CATEGORY_BROWSABLE} and the Intent
* data scheme is "http" or "https".
*
* @param onlyWebSchemes When true, requires that the intent filter declare
* that it handles *only* http: or https: schemes. This is a requirement for
* the intent filter's domain linkage being verifiable.
* @hide
*/
public final boolean handlesWebUris(boolean onlyWebSchemes) {
// Require ACTION_VIEW, CATEGORY_BROWSEABLE, and at least one scheme
if (!hasAction(Intent.ACTION_VIEW)
|| !hasCategory(Intent.CATEGORY_BROWSABLE)
|| mDataSchemes == null
|| mDataSchemes.size() == 0) {
return false;
}
// Now allow only the schemes "http" and "https"
final int N = mDataSchemes.size();
for (int i = 0; i < N; i++) {
final String scheme = mDataSchemes.get(i);
final boolean isWebScheme =
SCHEME_HTTP.equals(scheme) || SCHEME_HTTPS.equals(scheme);
if (onlyWebSchemes) {
// If we're specifically trying to ensure that there are no non-web schemes
// declared in this filter, then if we ever see a non-http/https scheme then
// we know it's a failure.
if (!isWebScheme) {
return false;
}
} else {
// If we see any http/https scheme declaration in this case then the
// filter matches what we're looking for.
if (isWebScheme) {
return true;
}
}
}
// We get here if:
// 1) onlyWebSchemes and no non-web schemes were found, i.e success; or
// 2) !onlyWebSchemes and no http/https schemes were found, i.e. failure.
return onlyWebSchemes;
}看到handlesWebUris()处理Intent.CATEGORY_BROWSABLE。需要注意的是，这个方法会根据传入参数的不同调整判断的标准，如果传入true显然只会识别http/https uri。而从Android DeepLink官方文档上看，uri可以定义成非http/https。另一个方法handleAllWebDataURI()可以满足这个限制：
/**
* Return if this filter handle all HTTP or HTTPS data URI or not. This is the
* core check for whether a given activity qualifies as a "browser".
*
* @return True if the filter handle all HTTP or HTTPS data URI. False otherwise.
*
* This will check if:
*
* - either the Intent category is {@link android.content.Intent#CATEGORY_APP_BROWSER}
* - either the Intent action is {@link android.content.Intent#ACTION_VIEW} and
* the Intent category is {@link android.content.Intent#CATEGORY_BROWSABLE} and the Intent
* data scheme is "http" or "https" and that there is no specific host defined.
*
* @hide
*/
public final boolean handleAllWebDataURI() {
return hasCategory(Intent.CATEGORY_APP_BROWSER) ||
(handlesWebUris(false) && countDataAuthorities() == 0);
}继续看handleAllWebDataURI()在什么地方被使用，发现在PackageManagerService的内部类ActivityIntentResolver中，PackageManagerService.ActivityIntentResolver.newResult()：
@Override
protected ResolveInfo newResult(PackageParser.ActivityIntentInfo info,
int match, int userId) {
if (!sUserManager.exists(userId)) return null;
if (!mSettings.isEnabledAndMatchLPr(info.activity.info, mFlags, userId)) {
return null;
}
final PackageParser.Activity activity = info.activity;
PackageSetting ps = (PackageSetting) activity.owner.mExtras;
if (ps == null) {
return null;
}
ActivityInfo ai = PackageParser.generateActivityInfo(activity, mFlags,
ps.readUserState(userId), userId);
if (ai == null) {
return null;
}
final ResolveInfo res = new ResolveInfo();
res.activityInfo = ai;
if ((mFlags&PackageManager.GET_RESOLVED_FILTER) != 0) {
res.filter = info;
}
if (info != null) {
res.handleAllWebDataURI = info.handleAllWebDataURI();
}
res.priority = info.getPriority();
res.preferredOrder = activity.owner.mPreferredOrder;
//System.out.println("Result: " + res.activityInfo.className +
// " = " + res.priority);
res.match = match;
res.isDefault = info.hasDefault;
res.labelRes = info.labelRes;
res.nonLocalizedLabel = info.nonLocalizedLabel;
if (userNeedsBadging(userId)) {
res.noResourceId = true;
} else {
res.icon = info.icon;
}
res.iconResourceId = info.icon;
res.system = res.activityInfo.applicationInfo.isSystemApp();
return res;
}其中24行使用了PackageParser.ActivityIntentInfo.handleAllWebDataURI()，PackageParser.ActivityIntentInfo是IntentFilter的子类。
继续追一下newResult()这个方法的使用逻辑。
PackageManagerService.ActivityIntentResolver继承自IntentResolver：
frameworks/base/services/core/java/com/android/server/IntentResolver.java
这是一个抽象类，并且属于Android框架内部类，不属于public api。newResult()这个方法在IntentResolver中定义。
在PackageManagerService中并没有对PackageManagerService.ActivityIntentResolver.newResult()的显式使用，所以这应该是一个模板模式，newResult()在基类的主框架流程中被隐式使用，子类只要根据子集的需求override。
在IntentResolver中调用关系：
queryIntent()/queryIntentFromList() -> buildResolveList() -> newResult()
回到PMS，发现一条重要的调用链：
PackageManagerService.queryIntentActivities() ->
PackageManagerService.ActivityIntentResolver.queryIntent() ->
IntentResolver.queryIntent()
作为对外的接口，PackageManager.queryIntentActivities()的作用是根据给定的Intent参数寻找Activity。
在AMS中，有对于这个方法的调用：
ActivityStackSupervisor.realStartActivityLocked() ->
ActivityManagerService.startSetupActivityLocked() ->
PackageManagerService.queryIntentActivities()
查阅相关资料可知（譬如刘超《深入解析Android5.0系统》），ActivityStackSupervisor.realStartActivityLocked()这个方法是start Activity流程中的一环。由此引发上述分析中的对于Intent.CATEGORY_BROWSABLE的处理。