LiveData Observer触发旧响应导致Android登录逻辑异常问题

问题根源 #

当前登录流程的核心问题是LiveData粘性特性+复用同一个LiveData实例：

• Repository中复用成员变量mLivedata，导致第二次登录请求返回的LiveData仍持有首次登录成功的旧数据
• 每次点击登录按钮都创建新的ViewModel和Observer，重复监听加剧了旧数据的粘性传递，最终导致无论输入何种凭证都会触发旧的登录成功逻辑

1. 修复Repository：每次请求生成新的LiveData实例 #

移除对全局mLivedata的复用，每次登录请求都创建新的MutableLiveData，确保数据隔离：

override fun cLoginResponse(
    context: Context,
    username: String,
    password: String,
    progressCallback: ProgressCallback
): MutableLiveData<LoginPojo> {
    // 每次请求创建新的LiveData实例
    val loginLiveData = MutableLiveData<LoginPojo>()
    
    progressCallback.onStart()
    val apiServices = rInstance.getRetrofitInstance()?.create(ApiServices::class.java)
    val requestBody = LoginBody(username, password)

    val call = apiServices?.getUser(requestBody)

    try {
        call?.enqueue(object : Callback<LoginPojo> {
            override fun onResponse(
                call: Call<LoginPojo>,
                response: Response<LoginPojo>
            ) {
                progressCallback.onFinish()
                // 修正原代码笔误，使用response.body()获取接口返回数据
                val loginResult = response.body() ?: LoginPojo(access_token = "")
                loginLiveData.postValue(loginResult)
            }

            override fun onFailure(call: Call<LoginPojo>, t: Throwable) {
                progressCallback.onFinish()
                Toast.makeText(context, "Login Failed", Toast.LENGTH_SHORT).show()
                // 失败时返回空token对象，确保Observer能收到状态
                loginLiveData.postValue(LoginPojo(access_token = ""))
            }
        })
    } catch (e: Exception) {
        e.printStackTrace()
        progressCallback.onFinish()
        Toast.makeText(context, "Login Failed", Toast.LENGTH_SHORT).show()
        loginLiveData.postValue(LoginPojo(access_token = ""))
    }
    return loginLiveData
}

2. 优化ViewModel：统一管理请求状态 #

ViewModel只需初始化一次，避免每次点击都创建新实例，同时可选持有当前请求的LiveData：

class LoginViewModel(application: Application) : AndroidViewModel(application) {
    private val loginRepo = AuthRepoImpl.getInstance()
    // 持有当前登录请求的LiveData，方便Fragment统一观察
    var currentLoginLiveData: LiveData<LoginPojo>? = null

    fun loginResponse(
        context: Context,
        username: String,
        password: String,
        progressCallback: ProgressCallback
    ): LiveData<LoginPojo> {
        currentLoginLiveData = loginRepo.cLoginResponse(context, username, password, progressCallback)
        return currentLoginLiveData!!
    }
}

3. 优化Fragment：避免重复观察 #

ViewModel和Observer只初始化一次，点击登录仅触发请求，不重复创建监听：

private lateinit var loginViewModel: LoginViewModel

override fun onViewCreated(view: View, savedInstanceState: Bundle?) {
    super.onViewCreated(view, savedInstanceState)
    // 仅初始化一次ViewModel
    loginViewModel = ViewModelProvider(this)[LoginViewModel::class.java]
    
    // 仅注册一次Observer，通过viewLifecycleOwner管理生命周期
    loginViewModel.currentLoginLiveData?.observe(viewLifecycleOwner) { loginPojo ->
        val accessToken = loginPojo.access_token ?: ""
        if (accessToken.isNotEmpty()) {
            Navigation.findNavController(requireView()).navigate(R.id.homepage)
            requireActivity().finish()
        } else {
            Toast.makeText(requireContext(), "Failed", Toast.LENGTH_SHORT).show()
        }
    }
}

// 点击登录仅触发请求，不重复创建ViewModel和Observer
private fun loginCheck(email: String, pass: String) {
    loginViewModel.loginResponse(requireContext(), email, pass, object : ProgressCallback {
        val opusLoader = OpusProgressLoader(requireContext())
        override fun onStart() {
            opusLoader.customDialogFunctionality(isCancelable = true)
        }

        override fun onFinish() {
            opusLoader.customDialogFunctionality(isCancelable = false)
        }
    })
}

额外优化建议 #

• 可以使用SingleLiveEvent（自定义LiveData）替代普通MutableLiveData，彻底解决粘性数据问题，适合登录结果这类一次性事件
• 确保退出主页返回登录页时，LoginFragment的生命周期正常，避免内存泄漏

内容的提问来源于stack exchange，提问作者Asikur Rahman