I have a python server and about 10 android clients, using sockets. It is really important that when the server sends a message, all clients receive it at the same time (say 1/10th of a second of difference).
But the connection is over Wifi, and some devices get the message later than others, which gives a very messy result. I don't want to get the latency of every device because this is a very unreliable approach. I want something as accurate as possible.
For example, in FPS games, it is common to have a countdown at the start of the round, and every player can start playing at the same time. What kind of logic lies behind this?
As for what my code currently looks like:
I use a BufferedReader in android to read every line sent by the server. The server is a console application in which you can type a message, and when you press enter, every listed client receives it with a new thread for every client.
java method receiving messages:
private void readMessage() throws IOException {
String data;
while ((data = mBufferedReader.readLine()) != null) {
data = data.toUpperCase();
if (data.startsWith("POSITION")) {
String[] splitData = data.split("/");
Log.d(Constants.TAG, splitData[1]);
mMainActivity.setDevicePosition(Integer.parseInt(splitData[1]));
} else {
String message = data.substring(data.indexOf('/') + 1, data.length());
int devices = Integer.parseInt(data.substring(0, data.indexOf('/')));
if (message.length() >= devices) {
message += " ";
} else {
int difference = devices - message.length();
for (int i = 0; i < difference; i++) {
message += " ";
}
}
mMainActivity.printMessage(message);
}
}
}
python line :
for cl in clients_list:
start_new_thread(send_message_thread, (cl, message,))
Related
Multiple invocations to the jsonRequest() method from different Threads (Timer-1 and EDT) do interfere with each other and even one call returns the result of a previous invocation
My CodeNameOne application uses a background Thread (Timer-1) to retrieve and display data from a REST service every second and it allows the user to issue commands that also issue REST calls from the EDT thread.
private Map<String, Object> jsonRequest(String url, String body, String cmd, int timeoutMs) {
long startTs = System.currentTimeMillis();
try {
request = new ConnectionRequest();
request.setReadResponseForErrors(true);
// request.setTimeout(timeoutMs);
// Shai: Timeout in Codename One is currently limited to connection timeout and
// doesn't apply to read timeout so once a connection is made it will last
request.setHttpMethod(cmd);
request.setPost(cmd.equalsIgnoreCase("POST") || cmd.equalsIgnoreCase("PUT") || cmd.equalsIgnoreCase("PATCH"));
if (body != null) {
request.addRequestHeader("Accept", "application/json");
request.setContentType("application/json");
request.setRequestBody(body);
request.setWriteRequest(true);
}
request.setUrl(url);
NetworkManager.getInstance().addToQueueAndWait(request);
long duration = System.currentTimeMillis() - startTs;
Log.p(cmd + ": " + url + " " + duration + " ms");
if (request.getResponseCode() >= 400 || request.getResponseData() == null) {
Log.p("responseCode=" + request.getResponseCode() + " responseData=" + request.getResponseData());
return null;
}
Log.p(cmd + ": " + url + " " + new String(request.getResponseData()));
Map<String, Object> result = new JSONParser().parseJSON(new InputStreamReader(new ByteArrayInputStream(request.getResponseData()), "UTF-8"));
return result;
} catch (Exception e) {
problemHandler.handle(cmd, url, e);
}
return null;
}
Actually result of multiple invocations get mixed up.
I would expect that each call to addToQueueAndWait() waits for the right result and returns just when the result is there.
I observed this problem to happen much more often on Android than on iOS or the simulator
I doubt that's what you are seeing. I see that request is defined in the class level as a variable so I'm guessing you are seeing a typical race condition where the request variable gets replaced while one is sent and by the time you reach the parsing it's a different object.
There is no need to use a thread for polling as networking already runs on a separate thread (or more this is usually determined in the init(Object) method).
I would suggest using a timer for a single invocation that's invoked after the response finishes.
A better approach would be websockets though: https://www.codenameone.com/blog/introducing-codename-one-websocket-support.html
With websockets the server can push out an update notification. This will save you the need to constantly poll the server. It saves on device battery life and resources on server/device.
I have probably a simple question but has me stumped. For my Android Wear application, I have two sensors working (step counter and heartrate).The wear app then sends these values back to the mobile application. I am sending them using the Message API. My stepcount sendMessage() and heartrate sendMessage() method look the same. Here is my heartrate sendMessage method.
private void sendMessageToHandheld(final String message) {
if (mGoogleApiClient == null)
return;
Log.d(LOG_TAG,"sending a message to handheld: "+message);
// use the api client to send the heartbeat value to our handheld
final PendingResult<NodeApi.GetConnectedNodesResult> nodes = Wearable.NodeApi.getConnectedNodes(mGoogleApiClient);
nodes.setResultCallback(new ResultCallback<NodeApi.GetConnectedNodesResult>() {
#Override
public void onResult(NodeApi.GetConnectedNodesResult result) {
final List<Node> nodes = result.getNodes();
if (nodes != null) {
for (int i=0; i<nodes.size(); i++) {
final Node node = nodes.get(i);
Wearable.MessageApi.sendMessage(mGoogleApiClient, node.getId(), message, bytes);
}
}
}
});
}
Problem is I am only using one messageReceived method on the mobile. So I cant differentiate from the step value coming in or the heartrate value coming in.
#Override
public void onMessageReceived(MessageEvent messageEvent) {
super.onMessageReceived(messageEvent);
Log.d(LOG_TAG, "received a message from wear: " + messageEvent.getPath());
if (messageEvent.getPath().contains("HEARTBEAT")){
// save the new heartbeat value
currentValue = Integer.parseInt(messageEvent.getPath());
if(handler!=null) {
// if a handler is registered, send the value as new message
Log.d(LOG_TAG, "received a heartbeat message from wear: " + messageEvent.getPath());
handler.sendEmptyMessage(currentValue);
}
}
else {
// save the new steps value
currentStepValue = Integer.parseInt(messageEvent.getPath());
if (handler != null) {
// if a handler is registered, send the value as new message
Log.d(LOG_TAG, "received a step message from wear: " + messageEvent.getPath());
handler.sendEmptyMessage(currentStepValue);
}
}
I tried passing in a byte array into the Heartrate sendMessage() and other strings as flags so that I could tell the values apart but no luck. Anyone know what the best way to go about this would be?
Any help would be appreciated.
Cheers
It seems you are sending the data inside the path attribute. This is not the correct use of this parameter.
Let's take a look at the MessageApi.sendMessage(GoogleApiClient client, String nodeId, String path, byte[] data method.
What you want to do is use String path to provide identifier for your message, for example in your case it would be step_counter and heartbeat. This way you can identify it on the other side, when you receive the message.
The sensor data should go into data field. You can put it raw there, but a better way is to create a DataMap and then serialize it into byte[]. This way you can enrich the data later easily.
I'm wondering if anybody can help me figure out what is causing the data I am sending to become corrupt.
My setup is currently an Arduino pro mini with a HM-10 bluetooth module connected (I have also tried HM-11 Module too) and an Android application to receive the bluetooth data.
Module setup: http://letsmakerobots.com/node/38009
If I send data with big enough intervals then the data is fine, but if I send the data continuously I see messages getting mixed up and lost. To test this I send "$0.1,0.2,0.3,0.4,0.5" to the Android application from the Arduino, sometimes the stream of data appears to send fine but other times it is really quite scrambled. Please see the below graphs that demonstrate this:
Good case:
Bad case:
Arduino code:
String inputString = ""; //Hold the incoming data.
boolean stringComplete = false; //Determines if the string is complete.
boolean realtime = false;
void setup()
{
Serial.begin(9600);
delay(500);
Serial.print("AT+START");
delay(500);
}
void loop()
{
if(stringComplete)
{
if(inputString.equals("rStart"))
{
Serial.println("$startACK");
realtime = true;
}
else if(inputString.equals("stop"))
{
Serial.println("$stopACK");
realtime = false;
}
else{
Serial.print(inputString);
}
inputString = "";
stringComplete = false;
}
if(realtime)
{
Serial.println("$0.1,0.2,0.3,0.4,0.5,0.6");
delay(10);
}
}
void serialEvent() {
while (Serial.available())
{
// get the new byte:
char inChar = (char)Serial.read();
if (inChar == '\n')
{
stringComplete = true;
}
else
{
inputString += inChar;
}
}
}
The Android side just receives the data and then parses it in an IntentService:
#Override
protected void onHandleIntent(Intent intent) {
//Incoming command.
String rawData = intent.getStringExtra(DataProcessingIntentService.REQUEST);
//Append our new data to our data helper.
Log.i(this.getClass().getName(), "Previous Raw: (" + DataProcessingHelper.getInstance().getData() + ")");
DataProcessingHelper.getInstance().appendData(rawData);
Log.i(this.getClass().getName(), "New Raw: (" + DataProcessingHelper.getInstance().getData() + ")");
commandStartIndex = DataProcessingHelper.getInstance().getData().indexOf("$");
commandEndIndex = DataProcessingHelper.getInstance().getData().indexOf("\n");
//Set this as the data starting point.
if(commandStartIndex != -1){
DataProcessingHelper.getInstance().offsetData(commandStartIndex);
}
//Ensure that a command has been found and that the end index is after the starting index.
if(commandStartIndex != -1 && commandEndIndex > commandStartIndex){
//Remove the command structure from the command.
command = DataProcessingHelper.getInstance().getData().substring(commandStartIndex+1, commandEndIndex-1); //Remove the \r\n end command.
DataProcessingHelper.getInstance().offsetData(commandEndIndex+1);
if(command.length() > 1){
//Split the data out of the comand.
splitData = command.split(",");
Log.i(this.getClass().getName(), "Broadcasting the processed data. (" + command + ")");
//Broadcast data.
Intent broadcastIntent = new Intent();
broadcastIntent.setAction(DataProcessingIntentService.RESPONSE);
broadcastIntent.addCategory(Intent.CATEGORY_DEFAULT);
broadcastIntent.putExtra(DataProcessingIntentService.RESPONSE, splitData);
sendBroadcast(broadcastIntent);
}else{
Log.e(this.getClass().getName(), "Command is less than 1 character long!");
}
}
}
Thank you for any help!
I have now figured out what was causing this problem. It appears that BLE only supports a maximum of 20 bytes per a transaction. The time between these transactions is different depending on what you are using. I'm currently using notifications which means that I can send 20 bytes every 7.5 milliseconds maximum. I have opted for 10 milliseconds to be safe. I will now need to look into breaking up packets into 20 bytes maximum to ensure no data corruption.
On Android, as soon as an NFC tag gets in proximity of the phone, the system delievers an intent to my application that contains an objects that allows me to read and write the NDEF message of this tag. Specifically, I can write to this tag as often as I want, while it's still in proxmity of the phone. The Java code below gives you an impression of what i mean:
Tag tag = intent.getParcelableExtra(NfcAdapter.EXTRA_TAG);
Ndef ndef = Ndef.get(tag);
ndef.writeNdefMessage(/* some NDEF data */); // first write
ndef.writeNdefMessage(/* some NDEF data */); // second write
// further writes
ndef.writeNdefMessage(/* some NDEF data */); // n-th write
Can I do the same on Windows Phone 8.x, or can I only carry out a single NDEF-message-write operation to the tag and then need to bring it into proximity again (move off RF field and come back into with the tag)?
I'm able to write to a tag more than once without separating it from the phone and tapping it again. See following code for example:
ProximityDevice device = ProximityDevice.GetDefault();
device.SubscribeForMessage("WriteableTag", WriteableTagHandler);
private void WriteableTagHandler(ProximityDevice sender, ProximityMessage message)
{
var message1= Encoding.Unicode.GetBytes("http://1stUrl.com");
var message2 = Encoding.Unicode.GetBytes("http://secondUrl.com");
sender.PublishBinaryMessage("WindowsUri:WriteTag", message1.AsBuffer(), (s, e) =>
{
s.StopPublishingMessage(e);
sender.PublishBinaryMessage("WindowsUri:WriteTag", message2.AsBuffer(), (se,r)=>
{
se.StopPublishingMessage(r);
});
});
}
EDIT:
I have just checked with two devices, and in fact, it is possible to write-read more than once without separating and tapping the phones again. See the example below, where one device sends 5 messages and the other receives all of them:
Device 1 (sender):
ProximityDevice device = ProximityDevice.GetDefault();
device.DeviceArrived += (e) =>
{
for (int i = 1; i <= 5; i++)
{
e.PublishMessage("Windows.mySubType", "message " + i.ToString(), (s, m) =>
{
s.StopPublishingMessage(m);
});
}
};
Device 2 (receiver):
ProximityDevice device = ProximityDevice.GetDefault();
device.SubscribeForMessage("Windows.mySubType", (s, e) =>
{
Dispatcher.BeginInvoke(() =>
{
MessageBox.Show(e.DataAsString);
});
});
I'm porting a game to use Google Play Game Services with multiplayer support. I'm using RealTimeSocket instead of realtime message because the game already has socket support.
To get the socket I call GamesClient.getRealTimeSocketForParticipant, and then I could get input and output streams as use it as a usual socket.
My problem is that if a device receives data before the call to getRealTimeSocketForParticipant, I will not be able to read this data. For instance:
Device A calls getRealTimeSocketForParticipant.
Device A sends "Hello".
Device B calls getRealTimeSocketForParticipant.
Device B receives nothing.
Device A sends "World".
Device B receives "World".
I have modified one of the example projects (ButtonClicker) and replicated the problem here. I have modified the code to use realtime socket, and modified the startGame method to this:
String mReceivedData = "";
byte mNextByteToSend = 0;
void startGame(boolean multiplayer)
{
mMultiplayer = multiplayer;
updateScoreDisplay();
switchToScreen(R.id.screen_game);
findViewById(R.id.button_click_me).setVisibility(View.VISIBLE);
GamesClient client = getGamesClient();
String myid = mActiveRoom.getParticipantId(client.getCurrentPlayerId());
ArrayList<String> ids = mActiveRoom.getParticipantIds();
String remoteId = null;
for(int i=0; i<ids.size(); i++)
{
String test = ids.get(i);
if( !test.equals(myid) )
{
remoteId = test;
break;
}
}
//One of devices should sleep in 5 seconds before start
if( myid.compareTo(remoteId) > 0 )
{
try
{
//The device that sleeps will loose the first bytes.
Log.d(TAG, "Sleeping in 5 seconds...");
Thread.sleep(5*1000);
}
catch(Exception e)
{
}
}
else
{
Log.d(TAG, "No sleep, getting socket now.");
}
try
{
final RealTimeSocket rts = client.getRealTimeSocketForParticipant(mRoomId, remoteId);
final InputStream inStream = rts.getInputStream();
final OutputStream outStream = rts.getOutputStream();
final TextView textView =((TextView) findViewById(R.id.score0));
//Thread.sleep(5*1000); Having a sleep here instead minimizes the risk to get the problem.
final Handler h = new Handler();
h.postDelayed(new Runnable()
{
#Override
public void run()
{
try
{
int byteToRead = inStream.available();
for(int i=0; i<byteToRead; i++)
{
mReceivedData += " " + inStream.read();
}
if( byteToRead > 0 )
{
Log.d(TAG, "Received data: " + mReceivedData);
textView.setText(mReceivedData);
}
Log.d(TAG, "Sending: " + mNextByteToSend);
outStream.write(mNextByteToSend);
mNextByteToSend++;
h.postDelayed(this, 1000);
}
catch(Exception e)
{
}
}
}, 1000);
}
catch(Exception e)
{
Log.e(TAG, "Some error: " + e.getMessage(), e);
}
}
The code ensures that one of the two devices sleeps 5 seconds before the call to getRealTimeSocketForParticipant. For the device that doesn't sleep the output will be something like:
No sleep, getting socket now.
Sending: 0
Sending: 1
Sending: 2
Sending: 3
Sending: 4
Received data: 0
Sending: 5
Received data: 0 1
Sending: 6
Received data: 0 1 2
That's expected, no data lost. But for the other device I get this:
Sleeping in 5 seconds...
Received data: 4
Sending: 0
Received data: 4 5
Sending: 1
Received data: 4 5 6
Sending: 2
Received data: 4 5 6 7
Sending: 3
The first bytes are lost. Is there anyway to avoid this?
If i'm understanding the API correctly, the messages exchanged through a real time socket are unrealiable, so you can't always have assurance that all players received all messages you sent. I couldn't find info about the network protocol used by RealTimeSocket, but I suspect it's UDP.
If that's really the case, I'm afraid there's little you can do short of implementing some sort of handshake yourself. Choose one device (ex.: the one with the lowest ID) to be the "synchronizer", and have it create a set with every other device. Send a message ("SYN") such as "where are you? x y z" (not literally, of course) every second, until the others respond "I'm here! (y)" ("ACK"). Remove from the set the devices that sent a response, until the set is empty. At this point, send everyone a "game's starting!" and go on.
Note that any of these messages can be lost: if the "ACK" is lost, next time the "SYN" is sent the device should answer again. If the "game's starting" message is lost, tough luck, the device will keep waiting until it receives a different message, at such point it should consider itself free to start (though delayed).
One last note: even if the underlying protocol is TCP, it's still not 100% reliable, no protocol is. See this question for more info, if you don't know this fact already.