aes67-ESP32-P4/components/spotify/spotify.cpp

#include "spotify.h"

#include <atomic>
#include <cstring>
#include <exception>
#include <deque>
#include <map>
#include <memory>
#include <mutex>
#include <string>

#include "BellLogger.h"
#include "CSpotContext.h"
#include "LoginBlob.h"
#include "PlaybackState.h"
#include "SpircHandler.h"
#include "TrackPlayer.h"
#include "esp_log.h"
#include "freertos/FreeRTOS.h"
#include "freertos/task.h"
#include "mdns.h"

extern "C" {
#include "aes67_cfg.h"
#include "aes67_web.h"
}

#define BODY_MAX 8192

static const char *TAG = "spotify";

static std::shared_ptr<cspot::LoginBlob> s_blob;     // guarded by s_mutex
static std::mutex s_mutex;
static std::string s_name;                              // advertised device name ("" = not advertised)
static std::atomic<bool> s_enabled{false};
static std::atomic<bool> s_stop{false};                 // end the running session
static std::atomic<bool> s_got_blob{false};
static std::atomic<bool> s_in_session{false};           // session() running (its tasks may exist)
static std::atomic<int> s_volume{65535};                // player volume in Spotify units (0..65535)
static std::atomic<const char *> s_state{"disabled"};
static std::atomic<uint64_t> s_pcm_bytes{0};
static std::shared_ptr<cspot::SpircHandler> s_handler;    // running session, else null
static spotify_pcm_cb_t s_pcm_cb;
static spotify_event_cb_t s_event_cb;
static spotify_pos_cb_t s_written, s_played;

// Track boundaries in the output buffer: when playback passes one, the new track is audible and
// Spotify must be told (notifyAudioReachedPlayback); without it the app's state stalls after the
// first track and it drops the device. Guarded by s_bound_mutex (data callback vs session loop).
struct Boundary {
    uint32_t pos;
    std::string id;
};
static std::mutex s_bound_mutex;
static std::deque<Boundary> s_bounds;
static std::string s_last_id;           // track of the most recent data callback
static std::atomic<bool> s_depleted{false};

static std::mutex s_cmd_mutex;
static std::deque<spotify_cmd_t> s_cmds;    // from spotify_command(), run by the session loop
static uint32_t s_cmd_queued;               // commands queued so far (s_cmd_mutex)
static std::atomic<uint32_t> s_cmd_done{0}; // commands run so far

static std::mutex s_track_mutex;
static cspot::TrackInfo s_track;        // audible track (TRACK_INFO), guarded by s_track_mutex

static void clear_boundaries(bool new_load)
{
    std::lock_guard<std::mutex> lock(s_bound_mutex);
    s_bounds.clear();
    if (new_load) {
        s_last_id.clear();              // the next data (new track) marks a boundary
    }
}

// Session loop, every <= 200 ms.
static void check_playback(const std::shared_ptr<cspot::SpircHandler> &handler)
{
    if (!s_written || !s_played) {
        return;
    }
    uint32_t played = s_played();
    std::string reached;
    {
        std::lock_guard<std::mutex> lock(s_bound_mutex);
        while (!s_bounds.empty() && (int32_t)(played - s_bounds.front().pos) >= 0) {
            reached = s_bounds.front().id;      // latest boundary passed wins
            s_bounds.pop_front();
        }
    }
    if (!reached.empty()) {
        handler->notifyAudioReachedPlayback(reached);
        ESP_LOGI(TAG, "track audible, Spotify notified");
    }
    if (s_depleted && played == s_written()) {
        s_depleted = false;
        handler->notifyAudioEnded();
        ESP_LOGI(TAG, "playlist ended, Spotify notified");
    }
}

extern "C" void spotify_suspend(bool suspend)
{
    if (!suspend) {
        spotify_apply();
        return;
    }
    s_enabled = false;
    s_stop = true;
    s_got_blob = false;
    s_state = "suspended";
    for (int i = 0; i < 30 && s_in_session; i++) {  // session loop exits within 200 ms, then disconnect
        vTaskDelay(pdMS_TO_TICKS(100));
    }
    ESP_LOGI(TAG, "suspended%s", s_in_session ? " (session still ending)" : "");
}

extern "C" void spotify_set_volume(int pct)
{
    pct = pct < 0 ? 0 : pct > 100 ? 100 : pct;
    int v = pct * 65535 / 100;
    if (v == s_volume) {
        return;
    }
    s_volume = v;
    auto h = s_handler;
    if (h) {
        h->setRemoteVolume(v);
    }
}

extern "C" bool spotify_command(spotify_cmd_t cmd)
{
    if (!s_handler) {
        return false;
    }
    uint32_t seq;
    {
        std::lock_guard<std::mutex> lock(s_cmd_mutex);
        s_cmds.push_back(cmd);
        seq = ++s_cmd_queued;
    }
    // Wait until it has run (the loop runs every <= 200 ms), so callers can report the new state.
    for (int i = 0; i < 50 && (int32_t)(s_cmd_done - seq) < 0 && s_handler; i++) {
        vTaskDelay(pdMS_TO_TICKS(20));
    }
    return true;
}

// Session loop: run queued commands on the task that also handles Spotify's frames.
static void run_commands(const std::shared_ptr<cspot::SpircHandler> &handler)
{
    std::deque<spotify_cmd_t> cmds;
    uint32_t seq;
    {
        std::lock_guard<std::mutex> lock(s_cmd_mutex);
        cmds.swap(s_cmds);
        seq = s_cmd_queued;
    }
    for (auto cmd : cmds) {
        switch (cmd) {
        case SPOTIFY_CMD_PLAY:
        case SPOTIFY_CMD_PAUSE:
            handler->setPause(cmd == SPOTIFY_CMD_PAUSE);    // notifies the app, sends PLAY_PAUSE
            break;
        case SPOTIFY_CMD_NEXT:
        case SPOTIFY_CMD_PREV:
            // The new track's load notifies the app and flushes our buffer (PLAYBACK_START).
            if (!(cmd == SPOTIFY_CMD_NEXT ? handler->nextSong() : handler->previousSong())) {
                ESP_LOGW(TAG, "no %s track", cmd == SPOTIFY_CMD_NEXT ? "next" : "previous");
            }
            break;
        }
        ESP_LOGI(TAG, "command %d", (int)cmd);
    }
    s_cmd_done = seq;
}

extern "C" bool spotify_session_active(void)
{
    return s_handler != nullptr;
}

extern "C" bool spotify_now_playing(spotify_track_t *t)
{
    auto h = s_handler;
    if (!h) {
        return false;
    }
    {
        std::lock_guard<std::mutex> lock(s_track_mutex);
        strlcpy(t->artist, s_track.artist.c_str(), sizeof(t->artist));
        strlcpy(t->title, s_track.name.c_str(), sizeof(t->title));
        strlcpy(t->album, s_track.album.c_str(), sizeof(t->album));
        t->duration_ms = s_track.duration;
    }
    // Same position the app shows: last reported position, plus the time since while playing.
    const auto &st = h->getPlaybackState()->innerFrame.state;
    t->paused = st.status != PlayStatus_kPlayStatusPlay;
    uint64_t pos = st.position_ms;
    if (!t->paused) {
        pos += h->getContext()->timeProvider->getSyncedTimestamp() - st.position_measured_at;
    }
    t->position_ms = t->duration_ms && pos > t->duration_ms ? t->duration_ms : (uint32_t)pos;
    return true;
}

const char *spotify_state(void)
{
    return s_state.load();
}

/* ----- cspot/bell logging -> esp_log (UART + syslog) ----- */

// cspot logs through bell::bellGlobalLogger; without one, CSPOT_LOG dereferences NULL.
class EspLogger : public bell::AbstractLogger {
    static void out(esp_log_level_t lvl, const std::string &file, int line, const std::string &sub,
                    const char *fmt, va_list ap)
    {
        char msg[256];
        vsnprintf(msg, sizeof(msg), fmt, ap);
        // CDN URLs carry signed tokens: keep them out of the log (syslog may leave the box).
        char *q = strstr(msg, "?__token__=");
        if (q) {
            strcpy(q, "?__token__=...");
        }
        size_t slash = file.find_last_of('/');
        const char *base = file.c_str() + (slash == std::string::npos ? 0 : slash + 1);
        ESP_LOG_LEVEL(lvl, sub.empty() ? "cspot" : sub.c_str(), "%s:%d: %s", base, line, msg);
    }

public:
    void debug(std::string file, int line, std::string sub, const char *fmt, ...) override
    {
        va_list ap;
        va_start(ap, fmt);
        out(ESP_LOG_DEBUG, file, line, sub, fmt, ap);
        va_end(ap);
    }
    void info(std::string file, int line, std::string sub, const char *fmt, ...) override
    {
        va_list ap;
        va_start(ap, fmt);
        out(ESP_LOG_INFO, file, line, sub, fmt, ap);
        va_end(ap);
    }
    void error(std::string file, int line, std::string sub, const char *fmt, ...) override
    {
        va_list ap;
        va_start(ap, fmt);
        out(ESP_LOG_ERROR, file, line, sub, fmt, ap);
        va_end(ap);
    }
};

/* ----- zeroconf (Spotify Connect discovery) on our httpd ----- */

static std::shared_ptr<cspot::LoginBlob> blob()
{
    std::lock_guard<std::mutex> lock(s_mutex);
    return s_blob;
}

static esp_err_t info_get(httpd_req_t *req)
{
    auto b = blob();
    if (!s_enabled || !b) {
        return httpd_resp_send_err(req, HTTPD_404_NOT_FOUND, "Spotify Connect is disabled");
    }
    std::string json = b->buildZeroconfInfo();
    httpd_resp_set_type(req, "application/json");
    return httpd_resp_send(req, json.data(), json.size());
}

static std::string url_decode(const char *s, size_t n)
{
    std::string out;
    for (size_t i = 0; i < n; i++) {
        if (s[i] == '+') {
            out += ' ';
        } else if (s[i] == '%' && i + 2 < n && isxdigit((unsigned char)s[i + 1]) && isxdigit((unsigned char)s[i + 2])) {
            char hex[3] = { s[i + 1], s[i + 2], 0 };
            out += (char)strtol(hex, nullptr, 16);
            i += 2;
        } else {
            out += s[i];
        }
    }
    return out;
}

// The Spotify app posts its login blob (action=addUser&userName=...&blob=...&clientKey=...).
static esp_err_t info_post(httpd_req_t *req)
{
    auto b = blob();
    if (!s_enabled || !b) {
        return httpd_resp_send_err(req, HTTPD_404_NOT_FOUND, "Spotify Connect is disabled");
    }
    if (req->content_len == 0 || req->content_len > BODY_MAX) {
        return httpd_resp_send_err(req, HTTPD_400_BAD_REQUEST, "bad body");
    }
    std::string body(req->content_len, '\0');
    size_t got = 0;
    while (got < body.size()) {
        int r = httpd_req_recv(req, body.data() + got, body.size() - got);
        if (r <= 0) {
            return ESP_FAIL;
        }
        got += r;
    }
    std::map<std::string, std::string> query;
    size_t pos = 0;
    while (pos < body.size()) {
        size_t amp = body.find('&', pos), end = amp == std::string::npos ? body.size() : amp;
        size_t eq = body.find('=', pos);
        if (eq != std::string::npos && eq < end) {
            query[url_decode(&body[pos], eq - pos)] = url_decode(&body[eq + 1], end - eq - 1);
        }
        pos = end + 1;
    }
    ESP_LOGI(TAG, "zeroconf %s from \"%s\"", query["action"].c_str(), query["userName"].c_str());
    b->loadZeroconfQuery(query);
    s_got_blob = true;
    static const char ok[] = "{\"status\":101,\"spotifyError\":0,\"statusString\":\"ERROR-OK\"}";
    httpd_resp_set_type(req, "application/json");
    return httpd_resp_send(req, ok, sizeof(ok) - 1);
}

/* ----- session ----- */

static const char *event_name(cspot::SpircHandler::EventType t)
{
    using E = cspot::SpircHandler::EventType;
    switch (t) {
    case E::PLAY_PAUSE: return "PLAY_PAUSE";
    case E::VOLUME: return "VOLUME";
    case E::TRACK_INFO: return "TRACK_INFO";
    case E::DISC: return "DISC";
    case E::NEXT: return "NEXT";
    case E::PREV: return "PREV";
    case E::SEEK: return "SEEK";
    case E::DEPLETED: return "DEPLETED";
    case E::FLUSH: return "FLUSH";
    case E::PLAYBACK_START: return "PLAYBACK_START";
    }
    return "?";
}

static void session(void)
{
    cJSON *src = cfg_get("source");
    std::string client_id = cJSON_GetObjectItemCaseSensitive(src, "spotify_client_id")->valuestring;
    std::string client_secret = cJSON_GetObjectItemCaseSensitive(src, "spotify_client_secret")->valuestring;
    int bitrate = (int)cJSON_GetObjectItemCaseSensitive(src, "spotify_bitrate")->valuedouble;
    cJSON_Delete(src);

    auto login = blob();
    auto ctx = cspot::Context::createFromBlob(login);
    ctx->config.clientId = client_id;
    ctx->config.clientSecret = client_secret;
    ctx->config.audioFormat = bitrate >= 320 ? AudioFormat_OGG_VORBIS_320
                              : bitrate >= 160 ? AudioFormat_OGG_VORBIS_160 : AudioFormat_OGG_VORBIS_96;
    // createFromBlob() starts at volume 0, which the app would show; report what we actually play.
    ctx->config.volume = s_volume;
    s_state = "connecting";
    ctx->session->connectWithRandomAp();
    auto token = ctx->session->authenticate(login);
    if (token.empty()) {
        s_state = "login failed";
        ESP_LOGE(TAG, "login failed (check the client ID/secret and that the account is Premium)");
        return;
    }
    ctx->session->startTask();
    auto handler = std::make_shared<cspot::SpircHandler>(ctx);
    handler->subscribeToMercury();
    // PCM (44.1 kHz stereo s16) to the player; returning fewer bytes makes cspot retry the rest,
    // so the player's blocking write paces decoding to playback speed.
    handler->getTrackPlayer()->setDataCallback([](uint8_t *data, size_t len, std::string_view id) -> size_t {
        if (s_written) {
            // First data of a new track: it becomes audible when playback reaches this position.
            std::lock_guard<std::mutex> lock(s_bound_mutex);
            if (id != s_last_id) {
                s_last_id = std::string(id);
                s_bounds.push_back({ s_written(), s_last_id });
            }
        }
        size_t frames = len / 4;
        size_t taken = s_pcm_cb ? s_pcm_cb(reinterpret_cast<const int16_t *>(data), frames) : frames;
        s_pcm_bytes += taken * 4;
        return taken * 4;
    });
    handler->setEventHandler([](std::unique_ptr<cspot::SpircHandler::Event> ev) {
        using E = cspot::SpircHandler::EventType;
        switch (ev->eventType) {
        case E::TRACK_INFO: {
            auto &ti = std::get<cspot::TrackInfo>(ev->data);
            ESP_LOGI(TAG, "track: %s - %s (%lu ms)", ti.artist.c_str(), ti.name.c_str(), (unsigned long)ti.duration);
            std::lock_guard<std::mutex> lock(s_track_mutex);
            s_track = ti;
            return;
        }
        case E::PLAY_PAUSE:
            if (s_event_cb) {
                s_event_cb(std::get<bool>(ev->data) ? SPOTIFY_EV_PAUSE : SPOTIFY_EV_PLAY, 0);
            }
            break;
        case E::FLUSH:
        case E::SEEK:
        case E::PLAYBACK_START:
            // The output buffer is dropped: pending boundaries are void. A new load (PLAYBACK_START)
            // gets a new boundary with its first data; a seek stays within the same track.
            clear_boundaries(ev->eventType == E::PLAYBACK_START);
            s_depleted = false;
            if (s_event_cb) {
                s_event_cb(SPOTIFY_EV_FLUSH, 0);
                if (ev->eventType == E::PLAYBACK_START) {
                    s_event_cb(SPOTIFY_EV_PLAY, 0);
                }
            }
            break;
        case E::VOLUME:
            s_volume = std::get<int>(ev->data);     // exact value from the app, no echo back
            if (s_event_cb) {
                s_event_cb(SPOTIFY_EV_VOLUME, std::get<int>(ev->data));
            }
            break;
        case E::DEPLETED:
            s_depleted = true;      // queue done: tell Spotify once the buffer has played out
            break;
        default:
            break;
        }
        ESP_LOGI(TAG, "event %s", event_name(ev->eventType));
    });
    s_handler = handler;
    s_state = "connected";
    ESP_LOGI(TAG, "connected as %s", ctx->config.username.c_str());
    clear_boundaries(true);
    s_depleted = false;
    {
        std::lock_guard<std::mutex> lock(s_cmd_mutex);
        s_cmds.clear();
        s_cmd_done = s_cmd_queued;
    }
    while (!s_stop) {
        ctx->session->handlePacket();       // waits at most 200 ms
        run_commands(handler);
        check_playback(handler);
    }
    // Stops the queue and player tasks (waits for them), then closes the connection.
    s_handler.reset();
    {
        std::lock_guard<std::mutex> lock(s_track_mutex);
        s_track = {};
    }
    handler->disconnect();
    ESP_LOGI(TAG, "session ended");
}

static void spotify_task(void *arg)
{
    while (true) {
        while (!s_enabled || !s_got_blob) {
            if (s_enabled) {
                s_state = "waiting for Spotify app";
            }
            vTaskDelay(pdMS_TO_TICKS(500));
        }
        s_got_blob = false;
        s_stop = false;
        s_in_session = true;
        try {
            session();
        } catch (const std::exception &e) {
            s_state = "error";
            ESP_LOGE(TAG, "session ended: %s", e.what());
        }
        s_handler.reset();
        s_in_session = false;
        vTaskDelay(pdMS_TO_TICKS(2000));
    }
}

extern "C" void spotify_apply(void)
{
    cJSON *src = cfg_get("source");
    std::string mode = cJSON_GetObjectItemCaseSensitive(src, "mode")->valuestring;
    std::string name = cJSON_GetObjectItemCaseSensitive(src, "spotify_name")->valuestring;
    bool creds = cJSON_GetObjectItemCaseSensitive(src, "spotify_client_id")->valuestring[0] &&
                 cJSON_GetObjectItemCaseSensitive(src, "spotify_client_secret")->valuestring[0];
    cJSON_Delete(src);
    bool want = (mode == "spotify" || mode == "auto") && creds;

    if (!want || name != s_name) {
        // Disable, or re-advertise under a new name: end any session, drop the old entry.
        s_enabled = false;
        s_stop = true;
        s_got_blob = false;
        if (!s_name.empty()) {
            mdns_service_remove("_spotify-connect", "_tcp");
            ESP_LOGI(TAG, "Spotify Connect \"%s\" withdrawn", s_name.c_str());
            s_name.clear();
        }
    }
    if (!want) {
        s_state = !creds ? "no client credentials" : "disabled (source mode)";
        if (!creds && (mode == "spotify" || mode == "auto")) {
            ESP_LOGW(TAG, "not started: set the Spotify client ID and secret (web UI, Source)");
        }
        return;
    }
    if (s_name.empty()) {
        {
            std::lock_guard<std::mutex> lock(s_mutex);
            s_blob = std::make_shared<cspot::LoginBlob>(name);
        }
        mdns_txt_item_t txt[] = { { "VERSION", "1.0" }, { "CPath", "/spotify_info" }, { "Stack", "SP" } };
        esp_err_t err = mdns_service_add(name.c_str(), "_spotify-connect", "_tcp", 80, txt, 3);
        s_name = name;
        ESP_LOGI(TAG, "Spotify Connect \"%s\" advertised (%s)", name.c_str(), esp_err_to_name(err));
    }
    s_enabled = true;
}

extern "C" esp_err_t spotify_init(spotify_pcm_cb_t pcm, spotify_event_cb_t event, spotify_pos_cb_t written,
                                  spotify_pos_cb_t played)
{
    s_pcm_cb = pcm;
    s_event_cb = event;
    s_written = written;
    s_played = played;
    if (!bell::bellGlobalLogger) {
        bell::bellGlobalLogger = new EspLogger();
    }
    static const httpd_uri_t get = { .uri = "/spotify_info", .method = HTTP_GET, .handler = info_get, .user_ctx = nullptr };
    static const httpd_uri_t post = { .uri = "/spotify_info", .method = HTTP_POST, .handler = info_post, .user_ctx = nullptr };
    web_register_uri(&get);
    web_register_uri(&post);
    if (xTaskCreate(spotify_task, "spotify", 32 * 1024, nullptr, 5, nullptr) != pdPASS) {
        return ESP_ERR_NO_MEM;
    }
    spotify_apply();
    return ESP_OK;
}