这是一个视频呼叫屏幕。它需要令牌和通道名称来工作，需要传递到初始化调用引擎。我将这些存储在一个数据类中，用作可变状态。
屏幕状态数据类

@Keep
data class CallScreenState(
    val callerId: Int? = null,
    val recieverId: Int? = null,
    val chatRoom: ChatRoom.Data? = null,
    val rtcToken: AgoraTokenResponse.TokenData? = null
)

而在viewmodel中初始化状态由以下代码实现：

var callScreenState by mutableStateOf(CallScreenState())

并且在聊天室和令牌API的成功响应上的视图模型中，用该代码更新状态。

callScreenState = callScreenState.copy(
                                chatRoom = chatRoom.data,//from response
                                rtcToken = token.data   //from response
                            )

从此处开始，预计将使用chatRoom和rtcToken的新更新值重新组成屏幕。

而在可组合

val screenState = remember {
    viewModel.callScreenState
}

此屏幕状态用于将值传递到init引擎

val mEngine = remember {
    initEngine(
        context,
        object : IRtcEngineEventHandler() {
            override fun onJoinChannelSuccess(channel: String?, uid: Int, elapsed: Int) {
                Timber.e("hhp-CallScreen onJoinChannelSuccess channel:$channel,uid:$uid,elapsed:$elapsed")
            }
            override fun onUserJoined(uid: Int, elapsed: Int) {
                Timber.e("hhp-CallScreen onUserJoined:$uid")
                val desiredUserList = remoteUserMap.toMutableMap()
                desiredUserList[uid] = null
                remoteUserMap = desiredUserList.toMap() as HashMap<Int, TextureView?>
            }
            override fun onUserOffline(uid: Int, reason: Int) {
                Timber.e("hhp-CallScreen onUserOffline:$uid")
                val desiredUserList = remoteUserMap.toMutableMap()
                desiredUserList.remove(uid)
                remoteUserMap = desiredUserList.toMap() as HashMap<Int, TextureView?>
            }
            override fun onNetworkQuality(uid: Int, txQuality: Int, rxQuality: Int) {
                Timber.e("hhp-CallScreen onNetworkQuality $uid $txQuality $rxQuality")
            }
        },
        screenState.chatRoom?.channelName ?: "",  //Not recomposing when value changes in viewmodel
        viewModel.userRole,
        token = screenState.rtcToken?.token ?: "" //Not recomposing when value changes in viewmodel
    )
}

这是initEngine函数的创建

fun initEngine(
    current: Context,
    eventHandler: IRtcEngineEventHandler,
    channelName: String,
    userRole: String,
    token: String
): RtcEngine =
    RtcEngine.create(current, BuildConfig.AGORA_APPID, eventHandler).apply {
        enableVideo()
        setChannelProfile(1)
        if (userRole == "Broadcaster") {
            setClientRole(1)
        } else {
            setClientRole(0)
        }
        //Expected to be recomposed when screen state value updated with new values
        joinChannel(token, channelName, "", 0)
    }

我知道在最开始的时候，屏幕状态中的通道名称和令牌在api调用之前是空的。一旦api用于获取令牌和聊天室成功，屏幕状态将从viewmodel更新，我希望initEngine功能将被再次调用，因为它应该重新组合。但事实并非如此。我是否遗漏了什么？如何在屏幕状态中的通道名称值发生变化时重新组合？