SalesforceAnalytics 源码分析

SalesforceAnalytics 可分为两个部分:

  • Event 存储
  • Event 上报

AnalyticsManager

AnalyticsManager 有三个属性:

key type comment
storeManager EventStoreManager 在 AnalyticsManager 的构造方法中初始化,只有 getter 没有 setter
deviceAppAttributes DeviceAppAttribute  
globalSequenceId int global sequence ID used by events

上面的三个属性在构造方法中进行初始化:

public AnalyticsManager(String uniqueId, 
                        Context context, 
                        String encryptionKey, 
                        DeviceAppAttributes deviceAppAttributes) { 
    storeManager = new EventStoreManager(uniqueId, context, encryptionKey);
    this.deviceAppAttributes = deviceAppAttributes;
    globalSequenceId = 0;
}

deviceAppAttributesglobalSequenceId 都可以通过 set 方法赋新值。

AnalyticsManage#reset 方法可以删除 storeManager 中的所有 Events:

public AnalyticsManager(String uniqueId, 
                        Context context, 
                        String encryptionKey, 
                        DeviceAppAttributes deviceAppAttributes) { 
    storeManager = new EventStoreManager(uniqueId, context, encryptionKey);
    this.deviceAppAttributes = deviceAppAttributes;
    globalSequenceId = 0;
}

DeviceAppAttribute

DeviceAppAttribute 包含了 App、OS、SDK、Device 的信息,它可以和 JSONObject 互相转换。

DeviceAppAttribute 的构造初始化位于应用层(SalesforceSDK 模块的 SalesforceAnalyticsManager.java 文件)。

key meaning value
appVersion App version packageInfo.versionName
appName App name 通过 SalesforceSDKManager 设置
osVersion OS version Build.VERSION.RELEASE
osName OS name android
nativeAppType App type Native(SalesforceSDKManager)
mobileSdkVersion Mobile SDK version SalesforceSDKManager.SDK_VERSION
deviceModel Device model Build.MODEL
deviceId Device ID  
clientId Client ID BootConfig 提供

Encryptor

Encryptor 是一个工具类,其用途包括 Encryption、Decryption、Hash。

初始化

从 Encryptor 的初始化开始分析:

Encrytor.init(getContext())

init 是一个静态方法,参数是一个 Context 实例。

init 方法首先检查文件系统是否支持加密:

private static boolean isFileSystemEncrypted;
public static boolean init(Context ctx) {
    // Checks if file system encryption is available and active.
    DevicePolicyManager devicePolicyManager = (DevicePolicyManager) ctx.getSystemService(Service.DEVICE_POLICY_SERVICE);
    isFileSystemEncrypted = devicePolicyManager.getStorageEncryptionStatus() == DevicePolicyManager.ENCRYPTION_STATUS_ACTIVE;
    // ...
}

然后尝试去获取最好的加密算法:

private static String bestCipherAvailable;
  
public static boolean init(Context ctx) {
  // Checks if file system encryption is available and active.
  // ...(omitted)
  // Make sure the cryptographic transformations we want to use are available.
  bestCipherAvailable = null;
  try {
    getBestCipher();
  } catch (GeneralSecurityException gex) {
    Log.e(TAG, "Security exception thrown", gex);
  }
  if (bestCipherAvailable == null) {
    return false;
  }
}

如果获取不到 cipher,则 init 失败,直接返回 falsegetBestCipher() 通过 javax.crypto.Cipher 去获取加密算法,使用 BC 作为算法提供者(Provider)。

BC(Bouncy Castle) is a collection of APIs used in cryptography.

Encryptor 并没有把获取到的 Cipher 实例保存下来,而是仅仅保存了 Cipher 的名称 - bestCipherAvailable,并且提供了一个默认值 —— PREFER_CIPHER_TRANSFORMATION

private static final String PREFER_CIPHER_TRANSFORMATION = "AES/CBC/PKCS5Padding";
private static String bestCipherAvailable;

getBestCipher() 会在多个地方被调用,所以 Encryptor 在初始化时先获取 best cipher 的名称 —— bestCipherAvailable,后续调用 getBestCipher 时会首先尝试使用缓存起来的 bestCipherAvailable 去获取 Cipher 实例。

Cipher cipher = null;
if (bestCipherAvailable != null) {
    return Cipher.getInstance(bestCipherAvailable, "BC");
}

如果 Cipher 实例获取失败,再尝试通过 PREFER_CIPHER_TRANSFORMATION 去获取:

try {
    cipher = Cipher.getInstance(PREFER_CIPHER_TRANSFORMATION, "BC");
    if (cipher != null) {
        bestCipherAvailable = PREFER_CIPHER_TRANSFORMATION;
    }
} catch (GeneralSecurityException gex1) {
    Log.e(TAG, "Preferred combo not available", gex1);
}

PREFER_CIPHER_TRANSFORMATION 获取成功之后,会把它赋值给 bestCipherAvailable。

通过以上代码分析得出如下结论:

bestCipherAvailabel 的赋值只有两种情况,要么是 “AES/CBC/PKCS5Padding”,要么是 null bestCipherAvailabel 作用是不是为了方便扩展?例如让用户自己指定算法。

加密

PREFER_CIPHER_TRANSFORMATION 指定的 AES(Advance Standard Encryption)是对称加密。 encrypt 方法除了需要传入 data 之外,还需要密钥 —— key:

static byte[] encrypt(byte[] data, byte[] key)

参数和返回值都是 byte 数组,encrypt 可分解为如下7个步骤:

  1. 调用 getBestCipher 方法获取 Cipher 实例
  2. 构造密钥规格——SecretKeySpec
  3. 构造初始化向量(16位)——IvParameterSpec
  4. 利用 SecretKeySpec 和 IvParameterSpec 来初始化 Cipher
  5. 利用 Cipher 对 data 进行加密(密文128位)
  6. 拼接 IV 和 加密后的 data 到一个大数据 result
  7. 返回 result
final Cipher cipher = getBestCipher();
final SecretKeySpec skeySpec = new SecretKeySpec(key, cipher.getAlgorithm());
 
// Generates a unique IV per encryption.
byte[] initVector = generateInitVector();
final IvParameterSpec ivSpec = new IvParameterSpec(initVector);
cipher.init(Cipher.ENCRYPT_MODE, skeySpec, ivSpec);
byte[] meat = cipher.doFinal(data);
 
// Prepends the IV to the encoded data (first 16 bytes / 128 bits).
byte[] result = new byte[initVector.length + meat.length];
System.arraycopy(initVector, 0, result, 0, initVector.length);
System.arraycopy(meat, 0, result, initVector.length, meat.length);
return result;

初始化向量(initVector)是利用方法 generateInitVector 生成的:

private static byte[] generateInitVector() throws NoSuchAlgorithmException, NoSuchProviderException {
    final SecureRandom random = SecureRandom.getInstance("SHA1PRNG");
    byte[] iv = new byte[16];
    random.nextBytes(iv);
    return iv;
}

名词解释:

SecureRandom

  • This class provides a cryptographically strong random number generator (RNG).

IV(Initialization Vector)

  • In cryptography, an initialization vector (IV) or starting variable (SV) is a fixed-size input to a cryptographic primitive that is typically required to be random or pseudorandom.

AES

  • AES is a variant of Rijndael which has a fixed block size of 128 bits, and a key size of 128, 192, or 256 bits

Cryptor 根据参数和返回值的不同还提供了两个基于 AES-256(256 是 key 的长度)的加密方法:

一、(String data, String key) => byte[]: base64

byte[] keyBytes = Base64.decode(key, Base64.DEFAULT);
byte[] dataBytes = data.getBytes(UTF8);
return Base64.encode(encrypt(dataBytes, keyBytes), Base64.DEFAULT);

二、(String data, String key) => String: US-ASCII

byte[] bytes = encryptBytes(data, key);
return new String(bytes, "US-ASCII");

解密

decrypt 方法是 encrypt 方法的逆,它的声明如下:

static byte[] decrypt(byte[] data, int offset, int length, byte[] key)

步骤一、首先获取 encrypt 方法中写入的 IV:

// Grabs the init vector prefix (first 16 bytes / 128 bits).
byte[] initVector = new byte[16];
System.arraycopy(data, offset, initVector, 0, initVector.length);

步骤二、然后获取密文——meat:

// Grabs the encrypted body after the init vector prefix.
int meatLen = length - initVector.length;
int meatOffset = offset + initVector.length;
byte[] meat = new byte[meatLen];
System.arraycopy(data, meatOffset, meat, 0, meatLen);

步骤三、获取 SecretKeySpec 和 IvParameterSpec 来初始化 Cipher(与 encrypt 相反,init 的 mode 是 Cipher.DECRYPT_MODE):

final Cipher cipher = getBestCipher();
final SecretKeySpec skeySpec = new SecretKeySpec(key, cipher.getAlgorithm());
final IvParameterSpec ivSpec = new IvParameterSpec(initVector);
cipher.init(Cipher.DECRYPT_MODE, skeySpec, ivSpec);

步骤四、cipher 开始对密文——meat 进行解密处理:

byte[] padded = cipher.doFinal(meat, 0, meatLen);
byte[] result = padded;

PREFER_CIPHER_TRANSFORMATION 定义的 Cipher 名称是 AES/CBC/PKCS5Padding,里面有一个 padding,应该是用于填充 block size。 那么 block size 是什么呢,可以再看一次 AES 的定义:

AES is a variant of Rijndael which has a fixed block size of 128 bits, and a key size of 128, 192, or 256 bits

也就是说,block size 是 AES 的长度(128bit/16byte),如果密文长度不足 128 位,需要根据 PKCS5Padding 来填充剩余的位数:

PKCS5Padding is interpreted as a synonym for PKCS7Padding in the cipher specification. It is simply a historical artifact, and rather than change it Sun decided to simply pretend the PKCS5Padding means the same as PKCS7Padding when applied to block ciphers with a blocksize greater than 8 bytes.

因此,解密之后的数据需要去掉填充值:

byte paddingValue = padded[padded.length - 1];
if (0 <= paddingValue) {
    if (paddingValue < (byte) 16) {
        byte compare = padded[padded.length - paddingValue];
        if (compare == paddingValue) {
            result = new byte[padded.length - paddingValue];
            System.arraycopy(padded, 0, result, 0, result.length);
        }
    }
}

最后、返回 result。 Crypto 基于上面的 decrypt 算法还提供了另外两个解密方法: 一、(byte[] data, String key) => String

byte[] keyBytes = Base64.decode(key, Base64.DEFAULT);
byte[] dataBytes = Base64.decode(data, Base64.DEFAULT);
 
// Decrypts with AES-256.
byte[] decryptedData = decrypt(dataBytes, 0, dataBytes.length, keyBytes);
return new String(decryptedData, 0, decryptedData.length, UTF8);

二、(String data, String key) => String

return decrypt(data.getBytes(), key);

哈希

Hashing 用于验证数据的完整性(Integrity),并且算法不可逆。常见的哈希算法有:SHA1、SHA2(SHA256)、SHA3、MD5。

Cryptor 使用了 HMAC SHA-256:

In cryptography, a Keyed-hash Message Authentication code (HMAC) is a specific type of Message Authentication Code (MAC) involving a cryptographic hash function and a secret cryptographic key.

使用 HMAC 还需要提供一个密钥 —— secret cryptographic key。

static String hash(String data, String key) {

步骤一、把 String 类型的 data 和 key 转换成 UTF8 编码的 byte 数组:

byte [] keyBytes = key.getBytes(UTF8);
byte [] dataBytes = data.getBytes(UTF8);

步骤二、获取 MAC(Message Authentication Code) 实例:

Mac sha = Mac.getInstance(MAC_TRANSFORMATION, "BC");

步骤三、使用 SecreteKeySpec 初始化 MAC 实例(sha):

SecretKeySpec keySpec = new SecretKeySpec(keyBytes, sha.getAlgorithm());
sha.init(keySpec);

步骤四、计算哈希值:

byte [] sig = sha.doFinal(dataBytes);

步骤五、将哈希值(byte 数据)进行 Base64 编码变转换成字符串:

Base64.encodeToString(sig, Base64.NO_WRAP);

android.util.Base64 提供了以下几种编解码 flag

flag meaning
CRLF Encoder flag bit to indicate lines should be terminated with a CRLF pair instead of just an LF. Has no effect if NO_WRAP is specified as well.
NO_WRAP Encoder flag bit to omit all line terminators (i.e., the output will be on one long line).
NO_PADDING Encoder flag bit to omit the padding ‘=’ characters at the end of the output (if any).
URL_SAFE Encoder/decoder flag bit to indicate using the “URL and filename safe” variant of Base64 (see RFC 3548 section 4) where - and _ are used in place of + and *. DEFAULT Encoder/Decoder flag, RFC 2045
NO_CLOSE Flag to pass to Base64OutputStream to indicate that it should not close the output stream it is wrapping when it itself is closed.

判断是否是 Base64 编码可通过如下代码:

public static boolean isBase64Encoded(String key) {
    try {
        Base64.decode(key, Base64.DEFAULT);
        return true;
    } catch (IllegalArgumentException e) {
        return false;
    }
}

EventStoreManager

EventStoreManager 用于将 Event 数据加密后存储在文件系统。它有如下存储规则:

  • 每一个 Event 对应一个 File
  • File 的 rootDir 是 context.getFilesDir()
  • File Name 的规则是 EventID + filenameSuffix
    • EventID 通过 UUID.randomUUID() 来生成(by InstrumentationEventBuilder#buildEvent)
    • filenameSuffix 时构造时传入
  • Event 存入 File 之前需要加密(Encryptor)
  • 从 File 读取 Event 之后需要解密(Encryptor)
  • File 最大数是 1000(maxEvents = 1000)

初始化

因为 Cryptor 是基于 AES-256 进行加解密,因此构造时需要传入 encryptionKey。

public EventStoreManager(String filenameSuffix, Context context, String encryptionKey) {
  this.filenameSuffix = filenameSuffix;
  this.context = context;
  this.encryptionKey = encryptionKey;
  fileFilter = new EventFileFilter(filenameSuffix);
  rootDir = context.getFilesDir();
}

更改 encryptionKey

encryptionKey 之后可以更改,但是更改之后所有的 Event File 都需要重新加密存储。

public void changeEncryptionKey(String oldKey, String newKey) {
    final List<InstrumentationEvent> storedEvents = fetchAllEvents();
    deleteAllEvents();
    encryptionKey = newKey;
    storeEvents(storedEvents);
}

存储 Event

public void storeEvent(InstrumentationEvent event) {
    if (event == null || TextUtils.isEmpty(event.toJson().toString())) {
        Log.d(TAG, "Invalid event");
        return;
    }
    if (!shouldStoreEvent()) {
        return;
    }
    final String filename = event.getEventId() + filenameSuffix;
    FileOutputStream outputStream;
    try {
        outputStream = context.openFileOutput(filename, Context.MODE_PRIVATE);
        outputStream.write(encrypt(event.toJson().toString()).getBytes());
        outputStream.close();
    } catch (Exception e) {
        Log.e(TAG, "Exception occurred while saving event to filesystem", e);
    }
}

获取 Event

private InstrumentationEvent fetchEvent(File file) {
    if (file == null || !file.exists()) {
        Log.e(TAG, "File does not exist");
        return null;
    }
    InstrumentationEvent event = null;
    String eventString = null;
    final StringBuilder json = new StringBuilder();
    try {
        final BufferedReader br = new BufferedReader(new FileReader(file));
        String line;
        while ((line = br.readLine()) != null) {
            json.append(line).append('\n');
        }
        br.close();
        eventString = decrypt(json.toString());
    } catch (Exception ex) {
        Log.e(TAG, "Exception occurred while attempting to read file contents", ex);
    }
    if (!TextUtils.isEmpty(eventString)) {
        try {
            final JSONObject jsonObject = new JSONObject(eventString);
            event = new InstrumentationEvent(jsonObject);
        } catch (JSONException e) {
            Log.e(TAG, "Exception occurred while attempting to convert to JSON", e);
        }
    }
    return event;
}

EventStoreManager 获取 rootDir 文件夹下 Event File 的方式比较巧妙:

private List<File> getAllFiles() {
    final List<File> files = new ArrayList<File>();
    final File[] listOfFiles = rootDir.listFiles();
    for (final File file : listOfFiles) {
        if (file != null && fileFilter.accept(rootDir, file.getName())) {
            files.add(file);
        }
    }
    return files;
}
  
private static class EventFileFilter implements FilenameFilter {
    private String fileSuffix;
    public EventFileFilter(String fileSuffix) {
        this.fileSuffix = fileSuffix;
    }
 
    @Override
    public boolean accept(File dir, String filename) {
        return filename != null && filename.endsWith(fileSuffix);
    }
}

InstrumentationEvent

InstrumentationEvent 表示一个埋点事件,其包含如下字段:

InstrumentEvent 也可以和 JSONObject 互相转换。

InstrumentationEventBuilder

InstrumentationEvent 属性比较多,通过 InstrumentationEventBuilder 可方便地构造 InstrumentationEvent。

buildEvent 需要从 AnalyticsManager 处获取属性值,所以 InstrumentationEvent 的构造需要传入 AnalyticsManager 实例。

private InstrumentationEventBuilder(AnalyticsManager analyticsManager, Context context) {
    this.analyticsManager = analyticsManager;
    this.context = context;
}

build 过程中首先通过 UUID 生成 eventId:

final String eventId = UUID.randomUUID().toString();

如果发现必填的字段没有值,会直接抛出异常 - EventBuilderException。

sequenceId 每次加一:

int sequenceId = analyticsManager.getGlobalSequenceId() + 1;
analyticsManager.setGlobalSequenceId(sequenceId);

EventBuilderHelper 位于上层(SalesforceSDK 模块),用于应用层代码的埋点。它提供了同步和异步两种方式:

// 同步
static void createAndStoreEventSync(final String name, final UserAccount userAccount,
                                    final String className, final JSONObject attributes)
// 异步
static void createAndStoreEvent(final String name, final UserAccount userAccount,
                                final String className, final JSONObject attributes)

异步埋点通过一个后台线程来执行:

private static final ExecutorService threadPool = Executors.newFixedThreadPool(2);

createAndStoreEvent 方法内部的实现如下:

threadPool.execute(new Runnable() {
    @Override
    public void run() {
        createAndStore(name, userAccount, className, attributes);
    }
});

不管是同步还是异步,最终都要通过 createAndStore 实现埋点,而 createAndStore 会通过 InstrumentationEventBuilder 来构造一个 Event,然后通过 EventStoreManager 保存到文件中。

createAndStore 的 create 需要提供如下参数:

属性
name 参数传入
startTime System.currentTimeMillis()
page { "context": className(参数传入)}
schemaType SchemaType.LightningInteraction
eventType EventType.system

store 通过如下调用链保存到文件:

SalesforceAnalyticsManager ==> AnalyticsManager ==> EventStoreManager#storeEvent

Publisher

AnalyticsManager 负责存储 Event,Publisher 负责上报 Event。

AnalyticsPublisher

Publisher 和 Transformer 一一对应,他们的映射关系存储在 SalesforceAnalyticsManager 的成员变量 remotes 中:

public class SalesforceAnalyticsManager {
  private Map<Class<? extends Transform>, Class<? extends AnalyticsPublisher>> remotes;
}

SalesforceAnalyticsManager 是 AnalyticsManager 的应用层,上层代码的埋点都要使用 SalesforceAnalyticsManager:

上报埋点数据必须要用到 networking,而 networking 属于 SalesforceSDKCore 模块,因此 Publisher 也要属于 SalesforceSDKCore 模块。

Publisher 的接口定义如下:

public interface AnalyticsPublisher {
  boolean publish(JSONArray events);
}

AILTNPublisher

AnalyticsPublisher 负责上报 events,但是 events 需要首先转换成一个 JSONArray。SalesforceSDKCore 模块提供了一个具体的实现:

public class AILTNPublisher implements AnalyticsPublisher

AILTNPublisher 对应着 AILTNTransformer,它的 publish 方法会构造出 JSONArray

{
  "logLines": [
    {
      "code": "ailtn",
      "data": {
        "schemaType": "",
        "payload": {
        }
      }
    },
    {
      "code": "ailtn",
      "data": {
        "schemaType": "",
        "payload": {
        }
      }
    }
  ]
}

其中 payload 属性装载的是经过 AILTNTransfromer#tranform 之后的 event 数据(一个 JSONArray)。

数组组装完成之后会发往服务端(REST API 是 /services/data/v39.0/connect/proxy/app-analytics-logging)。

第一步、因为 SalesforceSDK 支持多用户,所以我们要先获取当前用户的 RestClient:

// RestClient allows you to send authenticated HTTP requests to a force.com server.
final RestClient restClient = SalesforceSDKManager.getInstance().getClientManager().peekRestClient();

第二步、首先用之前组装的 logLines 构造一个 RequestBody(Media Type 是 application/json; charset=utf-8):

RequestBody.create(RestRequest.MEDIA_TYPE_JSON, body.toString())

第三步、用 gzip 进行压缩:

gzipCompressedBody(RequestBody.create(RestRequest.MEDIA_TYPE_JSON, body.toString()))

gzipCompressBody 的实现方式如下:

private RequestBody gzipCompressedBody(final RequestBody body) {
    return new RequestBody() {
        @Override
        public MediaType contentType() { return body.contentType(); }
 
        @Override
        public long contentLength() {
            return -1; // We don't know the compressed length in advance!
        }
 
        @Override
        public void writeTo(BufferedSink sink) throws IOException {
            final BufferedSink gzipSink = Okio.buffer(new GzipSink(sink));
            body.writeTo(gzipSink);
            gzipSink.close();
        }
    };
}

类似 Decorator 模式,把未压缩的 RequestBody 数据压缩之后返回一个新的 RequestBody。

注意:contentLength() 方法的返回值是 -1,因为压缩之后的长度在 gzipCompressBody 创建的 RequestBody 中来不及计算。

第四步、计算 Content-Length:

private RequestBody setContentLength(final RequestBody requestBody) throws IOException {
    final Buffer buffer = new Buffer();
    requestBody.writeTo(buffer);
    return new RequestBody() {
        @Override
        public MediaType contentType() {
            return requestBody.contentType();
        }
 
        @Override
        public long contentLength() {
            return buffer.size();
        }
 
        @Override
        public void writeTo(BufferedSink sink) throws IOException {
            sink.write(buffer.snapshot());
        }
    };
}

当然也可以通过 Interceptor 压缩并计算长度:

class GzipRequestInterceptor implements Interceptor {
    @Override
    public Response intercept(Chain chain) throws IOException {
        Request originalRequest = chain.request();
        if (originalRequest.body() == null || originalRequest.header("Content-Encoding") != null) {
            return chain.proceed(originalRequest);
        }
 
        Request compressedRequest = originalRequest.newBuilder()
            .header("Content-Encoding", "gzip")
            .method(originalRequest.method(), setContentLength(gzip(originalRequest.body())))
            .build();
        return chain.proceed(compressedRequest);
  }
}

第五步、构造 HTTP HEADER 参数:

final Map<String, String> requestHeaders = new HashMap<>();
requestHeaders.put(CONTENT_ENCODING, GZIP);
requestHeaders.put(CONTENT_LENGTH, Long.toString(requestBody.contentLength()));

第六步、构造 RestRequest 发送请求:

final RestRequest restRequest = new RestRequest(
        RestRequest.RestMethod.POST, apiPath, requestBody, requestHeaders);
restResponse = restClient.sendSync(restRequest);

AnalyticsPublisherService

AnalyticsPublisherService 继承自 IntentService,在 onHandleIntent 方法中通过调用 SalesforceAnalyticsManager#publishAllEvents 方法把埋点数据发往服务端。

/**
 * Handles the publish action in the provided background thread.
 */
private void handleActionPublish() {
    final UserAccount userAccount = UserAccountManager.getInstance().getCurrentUser();
    if (userAccount != null) {
        final SalesforceAnalyticsManager analyticsManager = SalesforceAnalyticsManager.getInstance(userAccount);
        analyticsManager.publishAllEvents();
    }
}

因为 SalesforceSDK 支持多用户,所以调用 publishAllEvents 之前先要获取当前用户(userAccount)的 SalesforceAnalyticsManager。 publishAllEvents 最终还是要调回 AnalyticsPublisher 的 publish 方法。不过,它会遍历 remotes,把原始的 Event 列表通过 Transformer 进行格式转换,然后调用对应 Publisher 的 publish 方法。

SalesforceAnalyticsManager.java

private Map<Class<? extends Transform>, Class<? extends AnalyticsPublisher>> remotes;

因为 remotes 的 key 和 value 都是 Class 类型,所以要通过反射来实例化它们:

Transform transformer = null;
try {
    transformer = transformClass.newInstance();
} catch (Exception e) {
    Log.e(TAG, "Exception thrown while instantiating class", e);
}
  
AnalyticsPublisher networkPublisher = null;
try {
    networkPublisher = remotes.get(transformClass).newInstance();
} catch (Exception e) {
    Log.e(TAG, "Exception thrown while instantiating class", e);
}

遍历完成,所有的 publisher 都请求成功之后,删除所有的 events:

if (success) {
    eventStoreManager.deleteEvents(eventsIds);
}

AnalyticsPublisherService 提供一个静态方法用于启动 Service:

public static void startActionPublish(Context context) {
    final Intent intent = new Intent(context, AnalyticsPublisherService.class);
    intent.setAction(ACTION_PUBLISH);
    context.startService(intent);
}

那么,何时启动 Service 呢?SalesforceAnalyticsManager 在初始化时会创建一个「定时任务」,这个定时任务会每隔一段时间去启动一次 AnalyticsPublisherService:

private static final int DEFAULT_PUBLISH_FREQUENCY_IN_HOURS = 8;
private static ScheduledFuture createPublishHandler() {
    final ScheduledExecutorService scheduler = Executors.newSingleThreadScheduledExecutor();
    final Runnable publishRunnable = new Runnable() {
        @Override
        public void run() {
            AnalyticsPublisherService.startActionPublish(
                    SalesforceSDKManager.getInstance().getAppContext());
        }
    };
    return scheduler.scheduleAtFixedRate(publishRunnable, 0, sPublishFrequencyInHours,
            TimeUnit.HOURS);
}

返回值用一个静态成员变量 —— sScheduler 保存。 默认间隔时间是8个小时,当然也可以由用户指定。

// Adds a handler for publishing if not already active.
if (!sPublishHandlerActive) {
    sScheduler = createPublishHandler();
    sPublishHandlerActive = true;
}

sPublishHandlerActivity 也是一个静态变量,用于标记定时任务有没有启动。 因为 sScheduler 是一个静态变量,所以所有的用户(UserAccount)会共享一个定时任务 如果用户改变了定时任务的执行周期,需要通过过 sScheduler 结束掉当前的任务,然后重新启动一个新任务

多用户管理

SalesforceAnalyticsManager 通过一个 map 来管理多用户实例:

private static Map<String, SalesforceAnalyticsManager> INSTANCES;
public static synchronized SalesforceAnalyticsManager getInstance(UserAccount account,
                                                                  String communityId) {
  // 無関係なスースコードを省略
  String uniqueId = account.getUserId();
  if (UserAccount.INTERNAL_COMMUNITY_ID.equals(communityId)) {
      communityId = null;
  }
 
  if (!TextUtils.isEmpty(communityId)) {
      uniqueId = uniqueId + communityId;
  }
 
  SalesforceAnalyticsManager instance;
  if (INSTANCES == null) {
      INSTANCES = new HashMap<>();
      instance = new SalesforceAnalyticsManager(account, communityId);
      INSTANCES.put(uniqueId, instance);
  } else {
      instance = INSTANCES.get(uniqueId);
  }
 
  if (instance == null) {
      instance = new SalesforceAnalyticsManager(account, communityId);
      INSTANCES.put(uniqueId, instance);
  }
}

Transform

Transform 会把一个通用的 Event 转换成特定的目标格式。

例如把 InstrumentationEvent 转换成 JSONObject

public interface Transform {
  public JSONObject transform(InstrumentationEvent event);
}

SalesforceAnalytics 定义了一个把 Event 转换成 AILTN 格式的 AILTNTransformer。

不知道 AILTN 是什么,网上也没找到资料。

更新时间:

留下评论