OpenHarmony 3.2 Beta多媒体系列——音视频播放框架

关注

转载 2022-11-22 10:25:23 1k 阅读 AI 写同款· GEO 优化 ›

OpenHarmony 3.2 Beta多媒体系列——音视频播放gstreamer 多媒体播放框架主要的实现在PlayerServer服务中，这个服务提供了媒体播放框架所需要的实现环境，继续跟踪代码分析发现，PlayerServer主要通过gstreamer适配层，对gstreamer进行调用。gstreamer属于更加具体的实现，所以本篇文章主要是分析PlayerServer通过适配层调用到gstreamer的过程。本篇主要分析了多媒体gstreamer的调用，涉及到从PlayerServer到gstreamer的整体流程。1. 简介。阅读详情

一、简介

媒体子系统为开发者提供一套接口，方便开发者使用系统的媒体资源，主要包含音视频开发、相机开发、流媒体开发等模块。每个模块都提供给上层应用对应的接口，本文会对音视频开发中的音视频播放框架做一个详细的介绍。

二、目录

foundation/multimedia/media_standard

├── frameworks                        #框架代码
│   ├── js
│   │   ├── player
│   ├── native
│   │   ├── player                    #native实现
│   └── videodisplaymanager         #显示管理
│       ├── include
│       └── src
├── interfaces
│   ├── inner_api                     #内部接口
│   │   └── native
│   └── kits                          #外部JS接口
├── sa_profile                        #服务配置文件
└── services
    ├── engine                        #engine代码
    │   └── gstreamer
    ├── etc                           #服务配置文件
    ├── include                       #头文件
    └── services
        ├── sa_media                  #media服务
        │   ├── client                #media客户端
        │   ├── ipc                   #media ipc调用
        │   └── server                #media服务端
        ├── factory                   #engine工厂
        └── player                    #player服务
           ├── client                 #player客户端
           ├── ipc                    #player ipc调用
           └── server                 #player服务端

三、播放的总体流程

四、Native接口使用

OpenHarmony系统中，音视频播放通过N-API接口提供给上层JS调用，N-API相当于是JS和Native之间的桥梁，在OpenHarmony源码中，提供了C++直接调用的音视频播放例子，在foundation/multimedia/player_framework/test/nativedemo/player目录中。

void PlayerDemo::RunCase(const string &path)
{
    player_ = OHOS::Media::PlayerFactory::CreatePlayer();
    if (player_ == nullptr) {
        cout << "player_ is null" << endl;
        return;
    }
    RegisterTable();
    std::shared_ptr<PlayerCallbackDemo> cb = std::make_shared<PlayerCallbackDemo>();
    cb->SetBufferingOut(SelectBufferingOut());

    int32_t ret = player_->SetPlayerCallback(cb);
    if (ret != 0) {
        cout << "SetPlayerCallback fail" << endl;
    }
    if (SelectSource(path) != 0) {
        cout << "SetSource fail" << endl;
        return;
    }
    sptr<Surface> producerSurface = nullptr;
    producerSurface = GetVideoSurface();
    if (producerSurface != nullptr) {
        ret = player_->SetVideoSurface(producerSurface);
        if (ret != 0) {
            cout << "SetVideoSurface fail" << endl;
        }
    }
    SetVideoScaleType();
    if (SelectRendererMode() != 0) {
        cout << "set renderer info fail" << endl;
    }
    ret = player_->PrepareAsync();
    if (ret !=  0) {
        cout << "PrepareAsync fail" << endl;
        return;
    }
    cout << "Enter your step:" << endl;
    DoNext();
}

首先根据RunCase可以大致了解一下播放音视频的主要流程，创建播放器，设置播放源，设置回调方法（包含播放过程中的多种状态的回调），设置播放显示的Surface，这些准备工作做好之后，需要调用播放器的PrepareASync方法，这个方法完成后，播放状态会变成Prepared状态，这时就可以调用播放器的play接口，进行音视频的播放了。

RegisterTable()方法中，将字符串和对应的方法映射到Map中，这样后续的DoNext会根据输入的命令，来决定播放器具体的操作。

void PlayerDemo::DoNext()
{
    std::string cmd;
    while (std::getline(std::cin, cmd)) {
        auto iter = playerTable_.find(cmd);
        if (iter != playerTable_.end()) {
            auto func = iter->second;
            if (func() != 0) {
                cout << "Operation error" << endl;
            }
            if (cmd.find("stop") != std::string::npos && dataSrc_ != nullptr) {
                dataSrc_->Reset();
            }
            continue;
        } else if (cmd.find("quit") != std::string::npos || cmd == "q") {
            break;
        } else {
            DoCmd(cmd);
            continue;
        }
    }
}

void PlayerDemo::RegisterTable()
{
    (void)playerTable_.emplace("prepare", std::bind(&Player::Prepare, player_));
    (void)playerTable_.emplace("prepareasync", std::bind(&Player::PrepareAsync, player_));
    (void)playerTable_.emplace("", std::bind(&Player::Play, player_)); // ENTER -> play
    (void)playerTable_.emplace("play", std::bind(&Player::Play, player_));
    (void)playerTable_.emplace("pause", std::bind(&Player::Pause, player_));
    (void)playerTable_.emplace("stop", std::bind(&Player::Stop, player_));
    (void)playerTable_.emplace("reset", std::bind(&Player::Reset, player_));
    (void)playerTable_.emplace("release", std::bind(&Player::Release, player_));
    (void)playerTable_.emplace("isplaying", std::bind(&PlayerDemo::GetPlaying, this));
    (void)playerTable_.emplace("isloop", std::bind(&PlayerDemo::GetLooping, this));
    (void)playerTable_.emplace("speed", std::bind(&PlayerDemo::GetPlaybackSpeed, this));
}

以上的DoNext方法中核心的代码是func()的调用，这个func就是之前注册进Map中字符串对应的方法，在RegisterTable方法中将空字符串""和"play"对绑定为Player::Play方法，默认不输入命令参数时，是播放操作。

五、调用流程

本段落主要针对媒体播放的框架层代码进行分析，所以在流程中涉及到了IPC调用相关的客户端和服务端，代码暂且分析到调用gstreamer引擎。首先Sample通过PlayerFactory创建了一个播放器实例（PlayerImpl对象），创建过程中调用Init函数。

int32_t PlayerImpl::Init()
{
    playerService_ = MediaServiceFactory::GetInstance().CreatePlayerService();
    CHECK_AND_RETURN_RET_LOG(playerService_ != nullptr, MSERR_UNKNOWN, "failed to create player service");
    return MSERR_OK;
}

MediaServiceFactory::GetInstance()返回的是MediaClient对象，所以CreateplayerService函数实际上是调用了MediaClient对应的方法。

std::shared_ptr<IPlayerService> MediaClient::CreatePlayerService()
{
    std::lock_guard<std::mutex> lock(mutex_);
    if (!IsAlived()) {
        MEDIA_LOGE("media service does not exist.");
        return nullptr;
    }

    sptr<IRemoteObject> object = mediaProxy_->GetSubSystemAbility(
        IStandardMediaService::MediaSystemAbility::MEDIA_PLAYER, listenerStub_->AsObject());
    CHECK_AND_RETURN_RET_LOG(object != nullptr, nullptr, "player proxy object is nullptr.");

    sptr<IStandardPlayerService> playerProxy = iface_cast<IStandardPlayerService>(object);
    CHECK_AND_RETURN_RET_LOG(playerProxy != nullptr, nullptr, "player proxy is nullptr.");

    std::shared_ptr<PlayerClient> player = PlayerClient::Create(playerProxy);
    CHECK_AND_RETURN_RET_LOG(player != nullptr, nullptr, "failed to create player client.");

    playerClientList_.push_back(player);
    return player;
}

这个方法中主要通过PlayerClient::Create(playerProxy)方法创建了PlayerClient实例，并且将该实例一层层向上传，最终传给了PlayerImpl的playerService_变量，后续对于播放器的操作，PlayerImpl都是通过调用PlayerClient实例实现的。

int32_t PlayerImpl::Play()
{
    CHECK_AND_RETURN_RET_LOG(playerService_ != nullptr, MSERR_INVALID_OPERATION, "player service does not exist..");
    MEDIA_LOGW("KPI-TRACE: PlayerImpl Play in");
    return playerService_->Play();
}

int32_t PlayerImpl::Prepare()
{
    CHECK_AND_RETURN_RET_LOG(playerService_ != nullptr, MSERR_INVALID_OPERATION, "player service does not exist..");
    MEDIA_LOGW("KPI-TRACE: PlayerImpl Prepare in");
    return playerService_->Prepare();
}

int32_t PlayerImpl::PrepareAsync()
{
    CHECK_AND_RETURN_RET_LOG(playerService_ != nullptr, MSERR_INVALID_OPERATION, "player service does not exist..");
    MEDIA_LOGW("KPI-TRACE: PlayerImpl PrepareAsync in");
    return playerService_->PrepareAsync();
}

对于PlayerImpl来说，playerService_指向的PlayerClient就是具体的实现，PlayerClient的实现是通过IPC的远程调用来实现的，具体地是通过IPC中的proxy端向远端服务发起远程调用请求。

我们以播放Play为例：

int32_t PlayerClient::Play()
{
    std::lock_guard<std::mutex> lock(mutex_);
    CHECK_AND_RETURN_RET_LOG(playerProxy_ != nullptr, MSERR_NO_MEMORY, "player service does not exist..");
    return playerProxy_->Play();
}

int32_t PlayerServiceProxy::Play()
{
    MessageParcel data;
    MessageParcel reply;
    MessageOption option;

    if (!data.WriteInterfaceToken(PlayerServiceProxy::GetDescriptor())) {
        MEDIA_LOGE("Failed to write descriptor");
        return MSERR_UNKNOWN;
    }

    int error = Remote()->SendRequest(PLAY, data, reply, option);
    if (error != MSERR_OK) {
        MEDIA_LOGE("Play failed, error: %{public}d", error);
        return error;
    }
    return reply.ReadInt32();
}

proxy端发送调用请求后，对应的Stub端会在PlayerServiceStub::OnRemoteRequest接收到请求，根据请求的参数进行对应的函数调用。播放操作对应的调用Stub的Play方法。

int32_t PlayerServiceStub::Play()
{
    MediaTrace Trace("binder::Play");
    CHECK_AND_RETURN_RET_LOG(playerServer_ != nullptr, MSERR_NO_MEMORY, "player server is nullptr");
    return playerServer_->Play();
}

这里最终是通过playerServer_调用Play函数。playerServer_在Stub初始化的时候通过PlayerServer::Create()方式来获取得到。也就是PlayerServer。

std::shared_ptr<IPlayerService> PlayerServer::Create()
{
    std::shared_ptr<PlayerServer> server = std::make_shared<PlayerServer>();
    CHECK_AND_RETURN_RET_LOG(server != nullptr, nullptr, "failed to new PlayerServer");

    (void)server->Init();
    return server;
}

最终我们的Play调用到了PlayerServer的Play()。在媒体播放的整个过程中会涉及到很多的状态，所以在Play中进行一些状态的判读后调用OnPlay方法。这个方法中发起了一个播放的任务。

int32_t PlayerServer::Play()
{
    std::lock_guard<std::mutex> lock(mutex_);

    if (lastOpStatus_ == PLAYER_PREPARED || lastOpStatus_ == PLAYER_PLAYBACK_COMPLETE ||
        lastOpStatus_ == PLAYER_PAUSED) {
        return OnPlay();
    } else {
        MEDIA_LOGE("Can not Play, currentState is %{public}s", GetStatusDescription(lastOpStatus_).c_str());
        return MSERR_INVALID_OPERATION;
    }
}

int32_t PlayerServer::OnPlay()
{
    auto playingTask = std::make_shared<TaskHandler<void>>([this]() {
        MediaTrace::TraceBegin("PlayerServer::Play", FAKE_POINTER(this));
        auto currState = std::static_pointer_cast<BaseState>(GetCurrState());
        (void)currState->Play();
    });

    int ret = taskMgr_.LaunchTask(playingTask, PlayerServerTaskType::STATE_CHANGE);
    CHECK_AND_RETURN_RET_LOG(ret == MSERR_OK, ret, "Play failed");

    lastOpStatus_ = PLAYER_STARTED;
    return MSERR_OK;
}

在播放任务中调用了PlayerServer::PreparedState::Play()

int32_t PlayerServer::PreparedState::Play()
{
    return server_.HandlePlay();
}

在Play里面直接调用PlayerServer的HandlePlay方法，HandlePlay方法通过playerEngine_调用到了gstreamer引擎，gstreamer是最终播放的实现。

int32_t PlayerServer::HandlePlay()
{
    int32_t ret = playerEngine_->Play();
    CHECK_AND_RETURN_RET_LOG(ret == MSERR_OK, MSERR_INVALID_OPERATION, "Engine Play Failed!");

    return MSERR_OK;
}