Refactor to modular server/client/middleware architecture #44

tlrobinson · 2020-01-21T17:10:47Z

Hello, great work on this project!

I'm not sure if you'll want to take this PR as a whole, or take ideas from it, or not at all, but I thought I'd start the conversation here.

I was looking at what you had and found it difficult to add features, so I ended up refactoring it to be more modular. It's now essentially a library of components that can be plugged into each other in your main.cpp according to your application's requirements.

Basically there are 3 main abstract classes, and a number of classes that inherit from them

(by "client" and "server" I mean they're running in the same process calling each other directly, not communicating over the network or anything)

DisasterServer
- just one class/instances of this, DisasterRadio, acts as a sort of message bus sending messages from one "client" to all the others
- has transmit method called by clients (also connect/disconnect/setup/loop)
- calls receive methods on the clients
DisasterClient
- there are many instances of this, for example one WebSocketClient instance for each client, and one LoRaClient for the LoRa radio (via LoRaLayer2
- has receive method called by server (also setup/loop)
- calls transmit method on the server
DisasterMiddleware which implements both DisasterServer and DisasterClient
- sits between the server and clients, implementing "middleware" logic that can be shared between clients
- has transmit and receive methods called by server and client it sits between (also connect/disconnect/setup/loop)
Also, DisasterHistory, a simple interface for classes that store a history of messages
- has record (called by HistoryRecorder client) and replay (called by HistoryReplay middleware) methods

You can connect clients to the server like this:

  radio->connect(new LoRaClient());

And you can chain calls to connect (sort of similar to Node.js' Stream's pipe method) to add functionality from middleware like this:

  radio->connect(new WelcomeMessage())
       ->connect(new HistoryReplay(history))
       ->connect(new Console())
       ->connect(new StreamClient(&Serial))

(this sets up the serial port to be a client which uses the WelcomeMessage, HistoryReplay and
Console middleware)

Here are the modules currently implemented:

DisasterServer:
- DisasterRadio: the main "server" that takes messages from clients and sends them to other clients
DisasterClient:
- LoRaClient: interfaces with LoRaLayer2
- WebSocketClient: WebSocket connections from the web app
- StreamClient: for Arduino Stream, currently used for serial console
- TCPClient: for a telnet-like server
- HistoryRecord: records messages to SD card (or bounded queue in memory)
- OLEDClient: displays messages on the screen (mostly for debugging purposes, but eventually I'd like to use this for a mobile Disaster Radio terminal of some kind)
- GPSClient: proof-of-concept, interfaces with a serial GPS module to beacon your current location periodically
DisasterMiddleware:
- WelcomeMessage: (very simple, in main.cpp) shows a welcome message to clients when they connect
- HistoryReplay: shows history to clients when they connect
- Console: implements a simple console with chat (similar to the web app) plus some /commands (I went with a more IRC-like syntax but you could easily implement a getopt version)
DisasterHistory:
- HistorySD records history to SD card
- HistoryMemory records history to a bounded queue in memory (default limit 10 messages)

Here's the complete "topology" I currently have in my main.cpp (but the default firmware would likely include less):

void setup() {
  // most of setup() omitted

  radio->connect(new LoRaClient());

  radio->connect(new HistoryRecord(history));

  WebSocketClient::startServer(&ws_server, [](WebSocketClient *ws_client) {
    radio->connect(new WelcomeMessage())
        ->connect(new HistoryReplay(history))
        ->connect(ws_client);
  });

  TCPClient::startServer(&tcp_server, [](TCPClient *tcp_client) {
    radio->connect(new WelcomeMessage())
        ->connect(new HistoryReplay(history))
        ->connect(new Console())
        ->connect(tcp_client);
  });

  radio->connect(new WelcomeMessage())
      ->connect(new HistoryReplay(history))
      ->connect(new Console())
      ->connect(new StreamClient(&Serial));

  radio->connect(new OLEDClient(&display, 0, STATUS_BAR_HEIGHT + 1));

  radio->connect(new GPSClient(&Serial1));
}

Other features I've implemented in main.cpp:

You can set a WIFI_SSID and WIFI_PASSWORD to try to connect to, and if it fails it falls back to AP mode (this is really useful for development so I don't have to keep switching my laptop's WiFi)
It will detect if the web app is on the SD card or in SPIFFs, so if you insert an empty SD card (for recording history, for example) it will still load the web app from SPIFFs.

Caveats / TODOs:

this is the most C++ I've written in a long time so I'm probably doing things wrong. I'm not sure if I got memory management correct.
I'm currently using Arduino String for messages because it was easy, but I'm aware of the pitfalls and it would probably be better to convert everything to C strings, buffers + length, or some message container object
I assume messages don't include binary data, and I think there are some places I assume messages don't include newlines (history recording). That's probably not a safe assumption.
I didn't implement message IDs correctly.

Let me know what you think!

paidforby · 2020-01-22T15:41:10Z

Wow, this is a crazy commit. Thanks for the contribution!

If I understand the code correctly, this doesn't change any of the basic functionality, it is merely a different way of structuring and interacting with the existing code, along with some new features.

Adds a number of features I (and others I'm sure) have really wanted, namely the OLED display, telnet client, logging of messages and connecting to a home network (haven't tested that yet, but I bet it's great for development).

It moves the messy websocket code into a separate file, which has been on my list for a long time.

I appreciate the use of header guards, something I've been meaning to add to LoRaLayer2.

Really like the idea of using an IRC style console, pretty clever. There isn't a great reason to have a getopt style console, it's just what I'm most familiar with.

Had some minor issues compiling with platformio. It wasn't able to find TinyGPS++.h. I had to add it to the lib_deps in platformio.ini as [email protected]

I noticed that you have to uncomment the correct board in build_flags, updated instructions should be added to README.

This points to an older commit of LoRaLayer2 and does not have the latest updates to chat app. I think this is due to a misunderstanding I addressed in sudomesh/LoRaLayer2#2. Those updates can easily be merged into this refactor.

Currently, makeEspArduino is not supported. I was able to get it to compile by adding the necessary libraries, downgrading LoRaLayer2, adding the correct build_flag, and renaming main.cpp to main.ino. However, after compiling, it produces a massive quantity of linker errors, complaining about multiple definitions and undefined references, for a small snippet,

/tmp/mkESP/main_ttgo-lora32-v1/main.cpp.o:(.bss.sdInitialized+0x0): multiple definition of `sdInitialized'
/tmp/mkESP/main_ttgo-lora32-v1/main_.cpp.o:(.bss.sdInitialized+0x0): first defined here
/tmp/mkESP/main_ttgo-lora32-v1/main.cpp.o:(.bss.loraInitialized+0x0): multiple definition of `loraInitialized'
/tmp/mkESP/main_ttgo-lora32-v1/main_.cpp.o:(.bss.loraInitialized+0x0): first defined here
/tmp/mkESP/main_ttgo-lora32-v1/main.cpp.o:(.bss.radio+0x0): multiple definition of `radio'
/tmp/mkESP/main_ttgo-lora32-v1/main_.cpp.o:(.bss.radio+0x0): first defined here
/tmp/mkESP/main_ttgo-lora32-v1/main.cpp.o:(.bss.history+0x0): multiple definition of `history'
/tmp/mkESP/main_ttgo-lora32-v1/main_.cpp.o:(.bss.history+0x0): first defined here
/tmp/mkESP/main_ttgo-lora32-v1/main.cpp.o:(.bss.macaddr+0x0): multiple definition of `macaddr'
/tmp/mkESP/main_ttgo-lora32-v1/main_.cpp.o:(.bss.macaddr+0x0): first defined here

I started seeing these types of errors previously when I dabbled with using C++ code in this repo. It may be that makeEspArduino itself needs to be patched to allow it to link this code. I will look into it.

When I first started writing this repo, the idea was to keep it as C-like as possible. I knew, when I refactored LL2 to use C++ classes, I was bending the rules to make more readable code. Given that we are working in an arduino environment, it is silly to fool ourselves into thinking that we aren't using C++. Ultimately, I wanted to head towards something like this, I doubt I would have done as good of a job.

My main concerns are readability and support. This is certainly more comprehensible on a macro-level, in that it adds structure to something that was previously an amorphous blob at best and totally structure-less at worst. I have some difficulty understanding it on a lower-level, but that may be because it's essentially new code. I'm also not terribly well-versed in C++, as evidenced by my shoddy work on LoRaLayer2. As I read through the code and really start to use it, I'm sure it will begin to make sense.

I would like to merge as much of this as possible. There are a few fixes and updates I need make before accepting the merge.

Add the TinyGPSPlus library required to compile
Update README with new instructions for building with platformio
Update the message ID to be a running counter of sent messages, as it was before.
Bring LoRaLayer2 and chat app up to latest, as to not lose the routing table display to chat app.
Get it working with makeEspArduino, as to not lose the support for that option.

Once merged, we can work on switching from String to buffer + lengths or structs for the messages between server and clients.

I'm very busy this week, so I may not get to working on this until the weekend. But I will try to post updates on my progress to this thread.

Thanks again for the time + effort that clearly went into this PR!

paidforby · 2020-01-23T15:17:44Z

I was able to successfully compile and link this refactor with makeEspArduino. Those linker errors were mostly caused by missing libraries. The only annoying change is that main.cpp can't be named main.cpp. Apparently, that causes some strange conflict within makeEspArduino and claims an undefined reference to setup() and loop().

I opened an issue in the makeEspArduino repo, but I don't expect anything to come of it.

@tlrobinson your options for renaming main.cpp are either main.ino, firmware.cpp, or anything better you can think of that isn't main.cpp.

tlrobinson · 2020-01-23T18:23:52Z

Wow, this is a crazy commit. Thanks for the contribution!

It's been fun!

If I understand the code correctly, this doesn't change any of the basic functionality, it is merely a different way of structuring and interacting with the existing code, along with some new features.

Correct, unless I missed anything (the message ID is the only thing I can think of)

Really like the idea of using an IRC style console, pretty clever. There isn't a great reason to have a getopt style console, it's just what I'm most familiar with.

Had some minor issues compiling with platformio. It wasn't able to find TinyGPS++.h. I had to add it to the lib_deps in platformio.ini as [email protected]

Sorry, I added that at the last minute, I'll make sure it's working on a clean checkout.

I noticed that you have to uncomment the correct board in build_flags, updated instructions should be added to README.

Yeah, though we could default to TTGO LoRa32 V2 since that's what most people seem to be using.

BTW we could take inspiration (and even code, it's Apache licensed) from Paxcounter which supports a bunch of boards. Each one has it's own hardware abstraction layer file: https://github.com/cyberman54/ESP32-Paxcounter/tree/master/src/hal

This points to an older commit of LoRaLayer2 and does not have the latest updates to chat app. I think this is due to a misunderstanding I addressed in sudomesh/LoRaLayer2#2. Those updates can easily be merged into this refactor.

Correct, I'll update it.

Currently, makeEspArduino is not supported. I was able to get it to compile by adding the necessary libraries, downgrading LoRaLayer2, adding the correct build_flag, and renaming main.cpp to main.ino. However, after compiling, it produces a massive quantity of linker errors, complaining about multiple definitions and undefined references, for a small snippet,

Personally I don't believe it's worth supporting two build systems, and I'm obviously partial to Platformio, but I'm happy to help you keep makeEspArduino working if needed. Maybe we could run both builds in CI.

When I first started writing this repo, the idea was to keep it as C-like as possible. I knew, when I refactored LL2 to use C++ classes, I was bending the rules to make more readable code. Given that we are working in an arduino environment, it is silly to fool ourselves into thinking that we aren't using C++. Ultimately, I wanted to head towards something like this, I doubt I would have done as good of a job.

It's probably possible to do a similar architecture in C, but C++ seemed more natural to me.

My main concerns are readability and support. This is certainly more comprehensible on a macro-level, in that it adds structure to something that was previously an amorphous blob at best and totally structure-less at worst. I have some difficulty understanding it on a lower-level, but that may be because it's essentially new code. I'm also not terribly well-versed in C++, as evidenced by my shoddy work on LoRaLayer2. As I read through the code and really start to use it, I'm sure it will begin to make sense.

I can walk through a simplified example:

DisasterRadio *radio = new DisasterRadio();
AsyncWebSocket ws_server("/ws");

void setup() {
  // most of setup omitted

  radio->connect(new LoRaClient());

  WebSocketClient::startServer(&ws_server, [](WebSocketClient *ws_client) {
    radio->connect(ws_client);
  });
}

void loop()
{
  radio->loop();
}

This creates an instance of DisasterRadio which maintains a list of connected clients https://github.com/sudomesh/disaster-radio/pull/44/files#diff-cef2c270f28a3f1bf4474071468e8755R18

When radio->connect() is called with a client it is added to that list (as well as a list of newly connected clients, connected, I should probably rename that new_clients): https://github.com/sudomesh/disaster-radio/pull/44/files#diff-ea8c5481dd4f3db1633095cbd97be1aeR5 It also sets the client's server to itself so the client can transmit messages to the server.

In this case we're initially connecting a LoRaClient and setting up a callback via WebSocketClient::startServer that receives WebSocketClients when WebSockets connections are made, and we connect those to radio as well.

In the main loop() we call radio->loop() which calls client->setup() on each client the first time after it's connected, then calls client->loop() on each client after that (these are obviously analogous to Arduino's setup and loop). Clients aren't required to do anything in their setup() or loop() (for example since WebSocketClient is asynchronous it doesn't, but LoRaClient calls LL2.popFromChatBuffer() in loop().

If a client has a message it wants to send it calls server->transmit(this, message), e.x. LoRaClient and WebSocketClient. DisasterRadio then broadcasts that to all other clients via their receive methods (but not the client that sent it, though the client could pass NULL as the first argument of transmit if it wanted it echoed back)

That's basically it.

DisasterMiddleware (not shown in this example) just sits between radio and clients in order to intercept or inject new messages. You can see the default implementations of transmit, receive, setup, and loop call the same method on their client (therefore radio->connect(new DisasterMiddleware())->connect(new LoRaClient()) would just be equivalent to radio->connect(new LoRaClient())). Subclasses of DisasterMiddleware implement one or more of those methods to add functionality, for example WelcomeMessage and HistoryReplay just send some messages to the downstream client in setup. Console is more complicated and intercepts messages in both directions.

I would like to merge as much of this as possible. There are a few fixes and updates I need make before accepting the merge.

Add the TinyGPSPlus library required to compile

Update README with new instructions for building with platformio

Update the message ID to be a running counter of sent messages, as it was before.

Bring LoRaLayer2 and chat app up to latest, as to not lose the routing table display to chat app.

Get it working with makeEspArduino, as to not lose the support for that option.

I'm also open to splitting this PR up if you prefer.

"refactor" would pretty much just include:
- DisasterClient
- DisasterServer
- DisasterRadio
- WebSocketClient
- LoRaClient
"new features" would include:
- DisasterMiddleware
- HistoryRecord
- HistoryReplay
- OLEDClient
- GPSClient
- StreamClient
- TCPClient
- DisasterHistory
- HistorySD
- HistoryMemory
- Console

I could even do separate branches for history, console+serial+TCP, OLED, and GPS

Once merged, we can work on switching from String to buffer + lengths or structs for the messages between server and clients.

Something like your wsMessage probably makes sense so each client doesn't need to parse out the message ID and type. Is data null-terminated?

paidforby · 2020-01-23T19:38:57Z

BTW we could take inspiration (and even code, it's Apache licensed) from Paxcounter which supports a bunch of boards. Each one has it's own hardware abstraction layer file: https://github.com/cyberman54/ESP32-Paxcounter/tree/master/src/hal

Yes, someone else brought this to my attention. It looks promising. I'm in favor of adapting it to our needs. I have a Pycom LoPy4 sitting here I need to to test. I may try getting it to work with that.

Personally I don't believe it's worth supporting two build systems, and I'm obviously partial to Platformio, but I'm happy to help you keep makeEspArduino working if needed. Maybe we could run both builds in CI.

I'm happy to keep supporting makeEspArduino, since I have continued to use that. I know the changes that need to be made to get it working and I will push them when I merge. I like the idea of doing some CI tests, has been on my "to do" list.

Fyi, apparently, main.cpp is not a good name for an esp32 arduino sketch as it conflicts with the real main.cpp. Not sure how platformio deals with this conflict, but makeEspArduino does not like it. plerup/makeEspArduino#134 (comment)

Thanks for the simplified example. That helped a lot. I'm pretty sure I have a grasp of how it works. I probably just need to dive in and write a new client to really learn.

I don't see the point of splitting this in to separate PRs or branches, I would rather just keep everything in one place.

Let me know if you plan on taking care of the LoRaLayer2 and message ID updates. Once those are done, I will fix the makeEspArduino builds and merge your commits.

We should also work on documenting the operation of this refactor somewhere in the README or wiki.

matthewrwright · 2020-01-23T22:48:37Z

Happy to either provide both a Heltec V1 and a Heltec V2 board, or run tests on both/make accessible via SSH

samuk · 2020-01-24T14:18:57Z

What's the correct syntax for platformio.ini? I'm trying this on Debian for the TTGO board.

[env]
platform = espressif32
board = ttgo-lora32-v1
framework = arduino
upload_port = /dev/ttyUSB0
monitor_port = /dev/ttyUSB0
monitor_speed = 115200
build_flags =
-DTTGO_LORA_V1
; -DTTGO_LORA_V2
; -DTTGO_TBEAM

====

I get Error: Please setup environments in platformio.ini file

I also tried -D TTGO_LORA_V1 (with space) but that didn't work either.

paidforby · 2020-01-24T14:25:24Z

@samuk this is working for me,

[platformio]
src_dir = firmware/esp32_ttgo
data_dir = web/static


[env:ttgo-lora32-v1]
platform = espressif32
board = ttgo-lora32-v1
framework = arduino
upload_port = /dev/ttyUSB0
monitor_port = /dev/ttyUSB0
monitor_speed = 115200
build_flags = -DTTGO_LORA_V2

lib_deps =
    AsyncTCP
    ESP Async [email protected]
    [email protected]
    https://github.com/sudomesh/LoRaLayer2#3b61485779c0cfdad58a785a3e5b912f0cec5b7b
    https://github.com/paidforby/AsyncSDServer#13375c6be978cb34180378ecf4042a3a4a1f5eab
    [email protected]
    ESP8266 and ESP32 OLED driver for SSD1306 displays

Maybe try putting the build flag on the same line? That's the only obvious difference I see.

paidforby · 2020-01-24T14:32:20Z

Oh also I just noticed the env is different. Try using [env:ttgo-lora32-v1].

samuk · 2020-01-24T16:40:06Z

That builds OK, but pio run -t uploadfs -t upload fails with

A fatal error occurred: Invalid head of packet (0x01)
*** [uploadfs] Error 2

I can run pio run -t uploadfs successfully and afterwards pio run -t upload and they both complete. The OLED display shows MAC address, I see a Wifi Network and can load the webpage but it just displays 'disaster radio' and no app.

Don't worry about trying to debug it, I can just wait for the release.

paidforby · 2020-01-24T16:45:59Z

Interesting. Sounds like an issue with flashing the web app. Try flash the firmware and fs from either master or from 0.1.1 release. And then re-flash just the firmware from the refactor. I'll try to reproduce that bug while working on this merge.

samuk · 2020-01-24T16:55:55Z

Yeah nice that worked.. Thanks

firmware/esp32_ttgo/DisasterServer.h

paidforby · 2020-01-25T19:12:36Z

firmware/esp32_ttgo/client/GPSClient.cpp

+    {
+        if (gps.location.isValid() && gps.location.age() < beacon_period)
+        {
+            server->transmit(this, String("c|<" + username + "> ") + gps.location.lat() + ", " + gps.location.lng());


Maybe this should be a different message type? g or p maybe. In case the another client wants to handle the information differently

Yeah, I agree, and it should probably be a little more structured. The GPS support was really just a proof of concept.

Maybe there's some inspiration to be taken from ham radio's APRS https://en.wikipedia.org/wiki/Automatic_Packet_Reporting_System

firmware/esp32_ttgo/client/LoRaClient.cpp

firmware/esp32_ttgo/client/StreamClient.h

paidforby · 2020-01-25T19:34:58Z

firmware/esp32_ttgo/client/WebSocketClient.cpp

+    uint16_t msg_id = 0;
+
+    unsigned char buf[2 + message.length() + 2] = {'\0'};
+    memcpy(buf, &msg_id, 2);


This is unnecessary. The msg_id does not need to be added by the WebSocket Client, it is already included in the message by the transmitting node. This code is analogous to the sendToWS() function from the original code, https://github.com/sudomesh/disaster-radio/blob/master/firmware/esp32_ttgo/main.ino#L88. Therefore it should only convert message from a String to an unsigned char.

After testing, I see that it doesn't work with out that phony msg_id I think the reason is that in converting the received message (raw bytes) to a String type you lose the fact that the msg_ID is actually a two-byte binary value, instead it is represented by it's equivalent ASCII character. Thus, by adding the 0 character to the beginning, the String is now the correct length to be accepted by the web app.

Maybe there is some sort of work around by parsing the String and converting it back in binary, but the correct way to fix this, which was already discussed in the thread, is to pass around structs of char arrays (or pointers to char arrays) as the messages between the server and clients.

I will leave the current bandage fix in for now. But it should be fixed ASAP as it could cause unexpected side effects.

You'll see LoRaClient and WebSocketClient are the only clients that strip out message ID before calling transmit() and add in dummy message IDs in receive(), since those were the only existing clients I had to port.

paidforby · 2020-01-25T19:37:55Z

firmware/esp32_ttgo/config.h

+
+// secrets in a separate file so we don't accidentally commit credentials
+#if __has_include("secrets.h")
+#include "secrets.h"


I'm going to add an untracked secrets.h file to the repo, with those commented out lines and a note that this is where you should add secret information.

firmware/esp32_ttgo/middleware/Console.cpp

platformio.ini

paidforby · 2020-01-25T20:01:49Z

firmware/esp32_ttgo/main.cpp

+  }
+  if (WiFi.status() == WL_CONNECTED)
+  {
+    ip = WiFi.localIP();


If it fails to connect to your home wifi network, it should set this vaiue to 192.168.4.1, instead of the local 0.0.0.0 which is of no value to the user.

paidforby · 2020-01-25T20:05:17Z

firmware/esp32_ttgo/main.cpp

+    WiFi.setHostname(HOST_NAME);
+    WiFi.softAP(ssid);
+
+    ip = WiFi.localIP();


I guess the above comment should be placed down here. Same goes, we should figure out a way to dynamically retrieve the real IP address of the node in AP mode, maybe it could be just be set in config.h? Or maybe if we create a preferences file of some sort, that could then be reconfigured from the web gui or console?

paidforby

I reviewed all the files in this merge and left a number of comments. Nothing jumps out as a major problem. I have a commit ready that will address most of my concerns. Let me know if there is anything you desperately want to add before merging. I will wait until tomorrow night (1/26) to add my commits and merge.

firmware/esp32_ttgo/main.cpp

paidforby · 2020-01-27T01:49:46Z

firmware/esp32_ttgo/client/LoRaClient.cpp

+        memcpy(&msg_id, packet.data, 2);
+        memcpy(msg, packet.data + 2, len - 2);
+
+        server->transmit(this, String(msg));


This message should not be converted to a string as it contains binary information and causes the msg_id issue in the WebSocketClient. Will be left in on merge, but should be fixed ASAP.

Yeah, I'm stripping out the first 2 bytes by doing + 2 in the second memcpy. As we discussed, I think changing "message" to be a struct containing the message ID (as uint16_t?), message type (as char?) and message payload (null terminated C string or bytes + length?) would be the right way to preserve the message ID.

Are you saying the payload (excluding message ID) can contain binary data? In that case bytes+length rather than C string would be required.

Yes, the payload can contain binary data. For example, the routiing table packets, which are sent from LoRaLayer2 to the WebSocket are entirely binary data

…s LoRaLayer2 to latest, re-renames main.cpp to main.ino

tlrobinson · 2020-01-27T05:18:24Z

Thanks for making the changes and merging this! Sorry I didn't have a chance to make the changes myself.

Refactor to server/client/middleware architecture

0a569e8

tlrobinson requested a review from paidforby January 21, 2020 17:10

tlrobinson changed the title ~~Refactor to be more modular~~ Refactor to modular server/client/middleware architecture Jan 21, 2020

tlrobinson mentioned this pull request Jan 21, 2020

Revert sending outgoing messages back to the incoming buffer sudomesh/LoRaLayer2#2

Closed

Missing TinyGPSPlus lib_dep + misc main.cpp

0b8e7d8

beegee-tokyo mentioned this pull request Jan 24, 2020

Support for Bluetooth LE #45

Closed

matthewrwright mentioned this pull request Jan 24, 2020

heltec_wifi_lora_32_V2 support #40

Closed

paidforby mentioned this pull request Jan 25, 2020

Make the Pins to Connect to LoRa Variables #46

Closed