[feat] send control commands via JSON

src/device_controller/device_controller.h

@@ -9,7 +9,11 @@
 
 #include <stdio.h>
 
+#include <nlohmann/json.hpp>
+#include <string>
+
 #include "display_info.h"
+using json = nlohmann::json;
 
 namespace crossdesk {
 
@@ -55,7 +59,7 @@ typedef struct {
   int* bottom;
 } HostInfo;
 
-typedef struct {
+struct RemoteAction {
   ControlType type;
   union {
     Mouse m;
@@ -64,7 +68,111 @@ typedef struct {
     bool a;
     int d;
   };
-} RemoteAction;
+
+  // parse
+  std::string to_json() const { return ToJson(*this); }
+
+  bool from_json(const std::string& json_str) {
+    RemoteAction temp;
+    if (!FromJson(json_str, temp)) return false;
+    *this = temp;
+    return true;
+  }
+
+  static std::string ToJson(const RemoteAction& a) {
+    json j;
+    j["type"] = a.type;
+    switch (a.type) {
+      case ControlType::mouse:
+        j["mouse"] = {
+            {"x", a.m.x}, {"y", a.m.y}, {"s", a.m.s}, {"flag", a.m.flag}};
+        break;
+      case ControlType::keyboard:
+        j["keyboard"] = {{"key_value", a.k.key_value}, {"flag", a.k.flag}};
+        break;
+      case ControlType::audio_capture:
+        j["audio_capture"] = a.a;
+        break;
+      case ControlType::display_id:
+        j["display_id"] = a.d;
+        break;
+      case ControlType::host_infomation: {
+        json displays = json::array();
+        for (size_t idx = 0; idx < a.i.display_num; idx++) {
+          displays.push_back(
+              {{"name", a.i.display_list ? a.i.display_list[idx] : ""},
+               {"left", a.i.left ? a.i.left[idx] : 0},
+               {"top", a.i.top ? a.i.top[idx] : 0},
+               {"right", a.i.right ? a.i.right[idx] : 0},
+               {"bottom", a.i.bottom ? a.i.bottom[idx] : 0}});
+        }
+
+        j["host_info"] = {{"host_name", a.i.host_name},
+                          {"display_num", a.i.display_num},
+                          {"displays", displays}};
+        break;
+      }
+    }
+    return j.dump();
+  }
+
+  static bool FromJson(const std::string& json_str, RemoteAction& out) {
+    try {
+      json j = json::parse(json_str);
+      out.type = (ControlType)j.at("type").get<int>();
+      switch (out.type) {
+        case ControlType::mouse:
+          out.m.x = j.at("mouse").at("x").get<float>();
+          out.m.y = j.at("mouse").at("y").get<float>();
+          out.m.s = j.at("mouse").at("s").get<int>();
+          out.m.flag = (MouseFlag)j.at("mouse").at("flag").get<int>();
+          break;
+        case ControlType::keyboard:
+          out.k.key_value = j.at("keyboard").at("key_value").get<size_t>();
+          out.k.flag = (KeyFlag)j.at("keyboard").at("flag").get<int>();
+          break;
+        case ControlType::audio_capture:
+          out.a = j.at("audio_capture").get<bool>();
+          break;
+        case ControlType::display_id:
+          out.d = j.at("display_id").get<int>();
+          break;
+        case ControlType::host_infomation: {
+          std::string host_name =
+              j.at("host_info").at("host_name").get<std::string>();
+          strncpy(out.i.host_name, host_name.c_str(), sizeof(out.i.host_name));
+          out.i.host_name[sizeof(out.i.host_name) - 1] = '\0';
+          out.i.host_name_size = host_name.size();
+
+          out.i.display_num = j.at("host_info").at("display_num").get<size_t>();
+          auto displays = j.at("host_info").at("displays");
+
+          out.i.display_list =
+              (char**)malloc(out.i.display_num * sizeof(char*));
+          out.i.left = (int*)malloc(out.i.display_num * sizeof(int));
+          out.i.top = (int*)malloc(out.i.display_num * sizeof(int));
+          out.i.right = (int*)malloc(out.i.display_num * sizeof(int));
+          out.i.bottom = (int*)malloc(out.i.display_num * sizeof(int));
+
+          for (size_t idx = 0; idx < out.i.display_num; idx++) {
+            std::string name = displays[idx].at("name").get<std::string>();
+            out.i.display_list[idx] = (char*)malloc(name.size() + 1);
+            strcpy(out.i.display_list[idx], name.c_str());
+            out.i.left[idx] = displays[idx].at("left").get<int>();
+            out.i.top[idx] = displays[idx].at("top").get<int>();
+            out.i.right[idx] = displays[idx].at("right").get<int>();
+            out.i.bottom[idx] = displays[idx].at("bottom").get<int>();
+          }
+          break;
+        }
+      }
+      return true;
+    } catch (const std::exception& e) {
+      printf("Failed to parse RemoteAction JSON: %s\n", e.what());
+      return false;
+    }
+  }
+};
 
 // int key_code, bool is_down
 typedef void (*OnKeyAction)(int, bool, void*);

src/gui/render_callback.cpp

@@ -28,8 +28,10 @@ int Render::SendKeyCommand(int key_code, bool is_down) {
         client_properties_.end()) {
       auto props = client_properties_[controlled_remote_id_];
       if (props->connection_status_ == ConnectionStatus::Connected) {
-        SendDataFrame(props->peer_, (const char*)&remote_action,
+        json j = remote_action.to_json();
-                      sizeof(remote_action), props->data_label_.c_str());
+        std::string msg = j.dump();
+        SendDataFrame(props->peer_, msg.c_str(), msg.size(),
+                      props->data_label_.c_str());
       }
     }
   }
@@ -94,8 +96,11 @@ int Render::ProcessMouseEvent(const SDL_Event& event) {
       if (props->control_bar_hovered_ || props->display_selectable_hovered_) {
         remote_action.m.flag = MouseFlag::move;
       }
-      SendDataFrame(props->peer_, (const char*)&remote_action,
+
-                    sizeof(remote_action), props->data_label_.c_str());
+      json j = remote_action.to_json();
+      std::string msg = j.dump();
+      SendDataFrame(props->peer_, msg.c_str(), msg.size(),
+                    props->data_label_.c_str());
     } else if (SDL_EVENT_MOUSE_WHEEL == event.type &&
                last_mouse_event.button.x >= props->stream_render_rect_.x &&
                last_mouse_event.button.x <= props->stream_render_rect_.x +
@@ -127,8 +132,10 @@ int Render::ProcessMouseEvent(const SDL_Event& event) {
           (float)(event.button.y - props->stream_render_rect_.y) /
           render_height;
 
-      SendDataFrame(props->peer_, (const char*)&remote_action,
+      json j = remote_action.to_json();
-                    sizeof(remote_action), props->data_label_.c_str());
+      std::string msg = j.dump();
+      SendDataFrame(props->peer_, msg.c_str(), msg.size(),
+                    props->data_label_.c_str());
     }
   }
 
@@ -279,64 +286,54 @@ void Render::OnReceiveDataBufferCb(const char* data, size_t size,
     return;
   }
 
+  std::string json_str(data, size);
   RemoteAction remote_action;
-  memcpy(&remote_action, data, size);
+
+  try {
+    remote_action.from_json(json_str);
+  } catch (const std::exception& e) {
+    LOG_ERROR("Failed to parse RemoteAction JSON: {}", e.what());
+    return;
+  }
 
   std::string remote_id(user_id, user_id_size);
   if (render->client_properties_.find(remote_id) !=
       render->client_properties_.end()) {
     // local
     auto props = render->client_properties_.find(remote_id)->second;
-    RemoteAction host_info;
+    if (remote_action.type == ControlType::host_infomation &&
-    if (DeserializeRemoteAction(data, size, host_info)) {
+        props->remote_host_name_.empty()) {
-      if (ControlType::host_infomation == host_info.type &&
-          props->remote_host_name_.empty()) {
-        props->remote_host_name_ =
-            std::string(host_info.i.host_name, host_info.i.host_name_size);
-        LOG_INFO("Remote hostname: [{}]", props->remote_host_name_);
-
-        for (int i = 0; i < host_info.i.display_num; i++) {
-          props->display_info_list_.push_back(DisplayInfo(
-              std::string(host_info.i.display_list[i]), host_info.i.left[i],
-              host_info.i.top[i], host_info.i.right[i], host_info.i.bottom[i]));
-          LOG_INFO("Remote display [{}:{}], bound [({}, {}) ({}, {})]", i + 1,
-                   props->display_info_list_[i].name,
-                   props->display_info_list_[i].left,
-                   props->display_info_list_[i].top,
-                   props->display_info_list_[i].right,
-                   props->display_info_list_[i].bottom);
-        }
-      }
-    } else {
       props->remote_host_name_ = std::string(remote_action.i.host_name,
                                              remote_action.i.host_name_size);
       LOG_INFO("Remote hostname: [{}]", props->remote_host_name_);
-      LOG_ERROR("No remote display detected");
+
+      for (int i = 0; i < remote_action.i.display_num; i++) {
+        props->display_info_list_.push_back(
+            DisplayInfo(remote_action.i.display_list[i],
+                        remote_action.i.left[i], remote_action.i.top[i],
+                        remote_action.i.right[i], remote_action.i.bottom[i]));
+      }
     }
-    FreeRemoteAction(host_info);
+    FreeRemoteAction(remote_action);
   } else {
     // remote
-    if (ControlType::mouse == remote_action.type && render->mouse_controller_) {
+    if (remote_action.type == ControlType::mouse && render->mouse_controller_) {
       render->mouse_controller_->SendMouseCommand(remote_action,
                                                   render->selected_display_);
-    } else if (ControlType::audio_capture == remote_action.type) {
+    } else if (remote_action.type == ControlType::audio_capture) {
-      if (remote_action.a) {
+      if (remote_action.a)
         render->StartSpeakerCapturer();
-        render->audio_capture_ = true;
+      else
-      } else {
         render->StopSpeakerCapturer();
-        render->audio_capture_ = false;
+    } else if (remote_action.type == ControlType::keyboard &&
-      }
-    } else if (ControlType::keyboard == remote_action.type &&
                render->keyboard_capturer_) {
       render->keyboard_capturer_->SendKeyboardCommand(
           (int)remote_action.k.key_value,
           remote_action.k.flag == KeyFlag::key_down);
-    } else if (ControlType::display_id == remote_action.type) {
+    } else if (remote_action.type == ControlType::display_id &&
-      if (render->screen_capturer_) {
+               render->screen_capturer_) {
-        render->selected_display_ = remote_action.d;
+      render->selected_display_ = remote_action.d;
-        render->screen_capturer_->SwitchTo(remote_action.d);
+      render->screen_capturer_->SwitchTo(remote_action.d);
-      }
     }
   }
 }

xmake.lua

@@ -23,6 +23,7 @@ end
 add_requires("spdlog 1.14.1", {system = false})
 add_requires("imgui v1.91.5-docking", {configs = {sdl3 = true, sdl3_renderer = true}})
 add_requires("openssl3 3.3.2", {system = false})
+add_requires("nlohmann_json 3.11.3")
 
 if is_os("windows") then
     add_requires("libyuv", "miniaudio 0.11.21")
@@ -45,7 +46,7 @@ elseif is_os("macosx") then
         "CoreMedia", "CoreVideo", "CoreAudio", "AudioToolbox")
 end
 
-add_packages("spdlog", "imgui")
+add_packages("spdlog", "imgui", "nlohmann_json")
 
 includes("submodules")