!7

sbool autoClose;
static int initialDelay = 0;
sbool infoBox, repeatEverything;

static int port;
static O onUtterance; // voidfunc(S)
sS botToSendUtterancesTo = "Voice Actions.";
static L<WebSocket> webSockets = synchroList();
sbool startRecognition;
static java.util.Timer stopper;
sS language = "en-US";

p {
  load('botToSendUtterancesTo);
  load('language);
  infoBox = repeatEverything = isMainProgram();
  NanoHTTPD.SOCKET_READ_TIMEOUT = 24*3600*1000; // not long enough - TODO: Fix in NanoHTTPD
  
  port = serveHttpWithWebSockets(9999, func(NanoHTTPD.IHTTPSession handshake) {
    WebSocket ws = new WebSocket(handshake) {
      protected void onPong(WebSocketFrame pongFrame) { print("pong"); }
      protected void onMessage(WebSocketFrame messageFrame) {
        cancelTimeout();
        fS s = messageFrame.getTextPayload();
        thread {
          pcallF(onUtterance, s);
          if (nempty(botToSendUtterancesTo))
            sendOpt(botToSendUtterancesTo, "User said: *", s);
        }
        if (repeatEverything) sendOpt("Mouth.", infoBoxAndReturn(switcheroo(s)));
        else infoBoxOrPrint("User said: " + s, infoBox);
      }
      protected void onClose(WebSocketFrame.CloseCode code, String reason, boolean initiatedByRemote) { webSockets.remove(this); }
      protected void onException(IOException e) { printStackTrace(e); }
    };
    if (startRecognition) {
      startRecognition = false;
      ws.send("start");
    }
    ret addAndReturn(webSockets, ws);
  });
  S url = print("http://localhost:" + port);
  //openPlatformBrowser(url);
  nohup("chromium-browser --app=" + url + "/popup");
  makeBot("Chrome Speech.");
}

html {
  if (neq(uri, "/popup"))
    ret hbody("Opening popup..." + hjavascript([[
      window.open('/popup', 'speech_recognizer', 'width=300,height=300,location=no');
      setTimeout(function() { window.close(); }, 10000);
    ]]));

  ret hhtml(hhead(htitle("Speech Recognizer")) + hbody(div(
    h3("Speech Recognizer")
    + loadJQuery()
    + hdiv("Results come here", id := 'results, style := "margin: 10px")
  + hjavascript([[
    var websocket;
    
    function stopRecognition() {
      recognition.stop();
      started = false;
      $("#btn").html("Start recognition");
      document.title = "Ready - Speech Recognizer";
    }
    
    function startRecognition() {
      recognition.start();
      started = true;
      $("#btn").html("Stop recognition");
      document.title = "Listening - Speech Recognizer";
    }
    
    function openWebSocket() {
      websocket = new WebSocket("ws://localhost:#PORT#/");
      websocket.onopen = function(event) {
        $("#btn").prop('disabled', false);
      };
    
      websocket.onmessage = function(event) {
        if (event.data == 'start' && !started) startOrStop();
        if (event.data == 'stop' && started) startOrStop();
        if (event.data.substring(0, 9) == 'language ')
          recognition.lang = event.data.substring(9);
      };
    
      if (#AUTOCLOSE#) websocket.onclose = function(event) {
        window.close();
      };
    }
    
    setTimeout(openWebSocket, #INITIALDELAY#);
      
    var recognition = new webkitSpeechRecognition();
    recognition.lang = "#LANGUAGE#";
    
    recognition.onerror = function(event) { 
      $("#results").html("Error: " + event.error);
      stopRecognition();
    }
    
    recognition.onresult = function(event) { 
      var result = event.results[0];
      var transcript = result[0].transcript;
      $("#results").html("Transcript: " + transcript);
      websocket.send(transcript);
      stopRecognition();
    }
    
    recognition.onnoresult = function(event) { 
      $("#results").html("-");
      stopRecognition();
    }
    
    var started = false;
    
    function startOrStop() {
      if (started) stopRecognition(); else {
        stopRecognition(); startRecognition();
      }
    }
    
    window.resizeTo(300, 300);
  ]]).replace("#PORT#", str(port)).replace("#AUTOCLOSE#", autoClose ? "true" : "false").replace("#INITIALDELAY#", str(initialDelay)).replace("#LANGUAGE#", language)
    + tag('button, "Start recognition", onclick := "startOrStop()", type := 'button, id := 'btn, disabled := 'disabled)
    //+ p(ahref("#", "Popup", onClick := "window.open('/', 'speech_recognizer', 'width=300,height=300,location=no'); return false;"));
  , style := "text-align: center"));
}

svoid startRecognition {
  L<WebSocket> l = cloneList(webSockets);
  if (empty(l)) startRecognition = true;
  else {
    //print("Starting recognition." + (l(l) > 1 ? "Weird: Have " + l(l) + " websockets" : ""));
    pcall {
      first(l).send("start");
    }
  }
}

svoid stopRecognition {
  if (startRecognition) startRecognition = false;
  if (nempty(webSockets)) pcall {
    first(webSockets).send("stop");
  }
}

answer {
  if "start recognition timeout *" {
    final int seconds = parseInt($1);
    startRecognition();
    stopper = timerOnce(toMS(seconds), f stopRecognition);
    ret "OK";
  }
  if "start recognition" { startRecognition(); ret "OK"; }
  if "stop recognition" { stopRecognition(); ret "OK"; }
  if "send to bot *" { setAndSave('botToSendUtterancesTo, $1); ret "OK"; }
  if "what bot are you sending to" ret quote(botToSendUtterancesTo);
  if "language *" {
    setAndSave('language, $1);
    pcall { if (nempty(webSockets)) first(webSockets).send("language " + $1); }
    ret "OK";
  }
}

svoid cancelTimeout {
  if (stopper != null) { stopper.cancel(); stopper = null; }
}