Release v0.6.49: Viber chat language profiles with local OCR

src/app.cpp

@@ -963,7 +963,7 @@ void App::refreshChatContext()
     HWND target = GetForegroundWindow();
     chatWindow = target;
     activeChatTitle = chatTitle(target);
-    activeChat = settings.chatCycles.empty() ? L"" : chatKey(processName(target), activeChatTitle);
+    activeChat = settings.chatCycles.empty() ? L"" : windowChatKey(target);
 }
 void App::switchChatLayout(bool reverse)
 {
@@ -974,10 +974,11 @@ void App::switchChatLayout(bool reverse)
     auto cycle = chatCycle(settings, key, layouts);
     auto available = layouts;
     resetInput();
-    enqueue([this, target, key, cycle, available, reverse]
+    enqueue([this, target, key, cycle, available, reverse, options = settings]
     {
         struct Reset { std::atomic<bool> &busy; ~Reset() { busy = false; } } reset{busy};
         auto valid = [&] { return GetForegroundWindow() == target && windowChatKey(target) == key; };
+        if (key.starts_with(L"viber\n") && recognizeViberChat(target, options) != key) return;
         if (!valid()) return;
         auto next = nextChatLayout(cycle, available, windowLayout(target), reverse);
         if (next) switchWindowLayout(target, next, valid);
@@ -1258,9 +1259,8 @@ void App::transform(int command, int snippet)
         version = inputVersion;
     }
     const auto requestedChat = config.chatCycles.empty() ? L"" : windowChatKey(target);
-    config.cycle = chatCycle(config, requestedChat, available);
     enqueue(
-        [this, target, recent, previous, config, available, version, command, snippet, attempt, manual, requestedAt, preDeferred]
+        [this, target, recent, previous, config, available, version, command, snippet, attempt, manual, requestedAt, preDeferred, requestedChat]() mutable
         {
             struct BusyReset
             {
@@ -1275,6 +1275,10 @@ void App::transform(int command, int snippet)
                         b = false;
                 }
             } reset{busy, window, preDeferred};
+            auto effectiveChat = requestedChat;
+            if (lower(processName(target)) == L"viber")
+                effectiveChat = recognizeViberChat(target, config);
+            config.cycle = chatCycle(config, effectiveChat, available);
             auto start = GetTickCount64();
             std::wstring reason = L"cancelled-before-selection";
             std::wstring text, result, expected, actual, verification = L"not-attempted";
@@ -2178,6 +2182,17 @@ LRESULT CALLBACK App::windowProc(HWND hwnd, UINT msg, WPARAM wp, LPARAM lp)
         return TRUE;
     }
     case WM_TIMER:
+        if (wp == 7)
+        {
+            static std::atomic<bool> pending{false};
+            auto target = GetForegroundWindow();
+            if (!app->settings.chatCycles.empty() && lower(processName(target)) == L"viber" && !pending.exchange(true))
+            {
+                auto options = app->settings;
+                app->enqueue([target, options] { recognizeViberChat(target, options); pending = false; });
+            }
+            return 0;
+        }
         if (wp == 6)
         {
             if (app->notificationRegistration && WaitForSingleObject(app->notificationRegistration, 0) == WAIT_OBJECT_0)
@@ -2318,6 +2333,7 @@ int App::execute()
     scheduleHistoryOcr(settings);
     addTray();
     SetTimer(window, 1, 60000, nullptr);
+    if (!smoke) SetTimer(window, 7, 750, nullptr);
     if (!smoke) SetTimer(window, 6, 1000, nullptr);
     SetThreadExecutionState(ES_CONTINUOUS |
                             (settings.awake ? (ES_DISPLAY_REQUIRED | ES_SYSTEM_REQUIRED) : 0));

src/chat_context.h

@@ -4,6 +4,8 @@
 
 namespace ks
 {
+std::wstring cachedViberChat(HWND);
+std::wstring recognizeViberChat(HWND, const Settings &);
 struct ChatSwitchHook
 {
     bool pressed[256]{};
@@ -64,6 +66,7 @@ inline std::wstring chatKey(std::wstring process, std::wstring title)
 }
 inline std::wstring windowChatKey(HWND window)
 {
+    if (window && lower(processName(window)) == L"viber") return cachedViberChat(window);
     return window ? chatKey(processName(window), chatTitle(window)) : L"";
 }
 inline std::vector<std::wstring> chatCycle(const Settings &settings, const std::wstring &key,

src/ocr.cpp

@@ -429,3 +429,4 @@ std::wstring recognizeText(const Clip &clip, const Settings &s)
     return wide(text.get());
 }
 } // namespace ks
+#include "viber_ocr.inc"

src/storage.cpp

@@ -67,7 +67,7 @@ void validateSettings(const Settings &s)
         auto separator = key.find(L'\n');
         if (separator == key.npos || separator + 1 == key.size() || key.size() > 1024 ||
             key.find(L'\n', separator + 1) != key.npos || key.find(L'\r') != key.npos ||
-            (key.substr(0, separator) != L"telegram" && key.substr(0, separator) != L"ms-teams"))
+            (key.substr(0, separator) != L"telegram" && key.substr(0, separator) != L"ms-teams" && key.substr(0, separator) != L"viber"))
             throw std::runtime_error("Invalid chat language profile");
         if (cycle.empty() || cycle.size() > 64) throw std::runtime_error("Select at least one chat language");
         std::vector<std::wstring> seen;

src/ui.cpp

@@ -1013,7 +1013,7 @@ LRESULT CALLBACK chatLanguagesProc(HWND w, UINT msg, WPARAM wp, LPARAM lp)
             {
                 int selected = (int)SendDlgItemMessageW(w, 710, LB_GETCURSEL, 0, 0);
                 if (selected < 0 || selected >= (int)v->keys.size())
-                    throw std::runtime_error("Open a Telegram or Teams chat, then reopen this dialog.");
+                    throw std::runtime_error("Open a Telegram, Teams or Viber chat, then reopen this dialog. Viber requires Tesseract.");
                 auto next = v->settings->draft;
                 if (id == 713) next.chatCycles.erase(v->keys[selected]);
                 else
@@ -1050,24 +1050,27 @@ void showChatLanguages(HWND parent, SettingsView *settings)
 {
     std::set<std::wstring> keys;
     for (const auto &[key, cycle] : settings->draft.chatCycles) keys.insert(key);
+    struct Context { std::set<std::wstring> &keys; const Settings &settings; } context{keys, settings->draft};
     EnumWindows([](HWND candidate, LPARAM context) -> BOOL
     {
         if (!IsWindowVisible(candidate)) return TRUE;
-        auto key = windowChatKey(candidate);
-        if (!key.empty()) ((std::set<std::wstring> *)context)->insert(std::move(key));
+        auto &data = *(Context *)context;
+        auto key = lower(processName(candidate)) == L"viber"
+            ? recognizeViberChat(candidate, data.settings) : windowChatKey(candidate);
+        if (!key.empty()) data.keys.insert(std::move(key));
         return TRUE;
-    }, (LPARAM)&keys);
+    }, (LPARAM)&context);
     HWND w = makeWindow(*settings->app, L"KSwitcher.ChatLanguages", L"Languages for a chat", chatLanguagesProc, 640, 524);
     auto *v = new ChatLanguagesView{parent, settings, {keys.begin(), keys.end()}};
     SetWindowLongPtrW(w, GWLP_HWNDPARENT, (LONG_PTR)parent);
     SetWindowLongPtrW(w, GWLP_USERDATA, (LONG_PTR)v);
-    label(w, L"Saved profiles and currently open Telegram / Teams chats", 18, 16, 604);
+    label(w, L"Telegram / Teams / Viber (experimental OCR) chat profiles", 18, 16, 604);
     control(w, L"LISTBOX", L"", WS_TABSTOP | WS_VSCROLL | LBS_NOTIFY | LBS_NOINTEGRALHEIGHT | WS_BORDER,
             710, 18, 44, 604, 140);
     for (const auto &key : v->keys)
     {
         auto split = key.find(L'\n');
-        auto name = std::wstring(key.starts_with(L"telegram\n") ? L"Telegram: " : L"Teams: ") + key.substr(split + 1);
+        auto name = std::wstring(key.starts_with(L"telegram\n") ? L"Telegram: " : key.starts_with(L"viber\n") ? L"Viber: " : L"Teams: ") + key.substr(split + 1);
         SendDlgItemMessageW(w, 710, LB_ADDSTRING, 0, (LPARAM)name.c_str());
     }
     control(w, L"STATIC", L"Open a chat, then reopen this dialog to add it.", 0, 712, 18, 194, 604, 26);
@@ -1986,7 +1989,7 @@ void showSettings(App &app, bool visible)
     label(w, L"Backspace / Delete clears it", 454, 270, 230);
     button(w, L"Re&fresh layouts", 253, 454, 322, 230);
     button(w, L"Languages for a &chat...", 259, 24, 378, 230);
-    label(w, L"Choose a separate set for each Telegram or Teams chat.", 272, 382, 412);
+    label(w, L"Chat languages: Telegram, Teams, Viber (OCR).", 272, 382, 412);
     page();
 
     label(w, L"Click a shortcut field and press the desired keys", 24, 76, 660);

src/update.h

@@ -2,7 +2,7 @@
 #include "platform.h"
 namespace ks
 {
-inline constexpr wchar_t AppVersion[] = L"0.6.48";
+inline constexpr wchar_t AppVersion[] = L"0.6.49";
 inline constexpr UINT UpdateMessage = WM_APP + 7;
 inline constexpr UINT UpdateCheckIntervalMs = 6 * 60 * 60 * 1000;
 inline constexpr UINT UpdateNotificationRetryMs = 5 * 60 * 1000;

src/viber_ocr.inc

@@ -0,0 +1,82 @@
+// Experimental Viber header OCR; never called on the keyboard hook thread.
+#include <uiautomation.h>
+#include <mutex>
+namespace ks {
+namespace {
+std::mutex viberMutex;
+HWND viberWindow{};
+std::wstring viberKey;
+ULONGLONG viberTime{};
+std::wstring elementClass(IUIAutomationElement *e) {
+ BSTR s{}; e->get_CurrentClassName(&s); std::wstring out=s?s:L""; SysFreeString(s); return out;
+}
+bool findViberHeader(IUIAutomationTreeWalker *walker,IUIAutomationElement *e,RECT &pane,RECT &search,int depth,int &budget) {
+ if(depth>5 || --budget<0) return false;
+ bool stack=elementClass(e).starts_with(L"StackView_");
+ Com<IUIAutomationElement> child;walker->GetFirstChildElement(e,child.out());
+ bool found=false,info=false;
+ while(child.p && --budget>0) {
+  if(stack) {
+   auto name=elementClass(child.p);
+   if(name.starts_with(L"SearchButton_")){child->get_CurrentBoundingRectangle(&search);found=true;}
+   if(name.starts_with(L"InfoButton_")) info=true;
+  } else if(findViberHeader(walker,child.p,pane,search,depth+1,budget)) return true;
+  Com<IUIAutomationElement> next;walker->GetNextSiblingElement(child.p,next.out());
+  child.p->Release();child.p=next.p;next.p=nullptr;
+ }
+ if(found && info){e->get_CurrentBoundingRectangle(&pane);return true;}
+ return false;
+}
+Pixels viberHeader(HWND window) {
+ if(!IsWindow(window)||IsIconic(window)) return {};
+ Com<IUIAutomation2> automation;
+ check(CoCreateInstance(CLSID_CUIAutomation8,nullptr,CLSCTX_INPROC_SERVER,IID_IUIAutomation2,(void**)automation.out()));
+ automation->put_ConnectionTimeout(200);automation->put_TransactionTimeout(200);
+ Com<IUIAutomationElement> root;check(automation->ElementFromHandle(window,root.out()));
+ Com<IUIAutomationTreeWalker> walker;check(automation->get_RawViewWalker(walker.out()));
+ RECT pane{},search{},rect{};int budget=160;
+ if(!findViberHeader(walker.p,root.p,pane,search,0,budget)||!GetWindowRect(window,&rect)) return {};
+ double scale=GetDpiForWindow(window)/96.0;
+ int x=pane.left+int(76*scale)-rect.left,y=search.top-int(12*scale)-rect.top;
+ int width=search.left-int(12*scale)-rect.left-x,height=int(28*scale);
+ int ww=rect.right-rect.left,wh=rect.bottom-rect.top;
+ if(width<30||height<10||x<0||y<0||x+width>ww||y+height>wh||uint64_t(ww)*wh>16000000) return {};
+ HDC dc=CreateCompatibleDC(nullptr);BITMAPINFO bi{};bi.bmiHeader.biSize=sizeof(BITMAPINFOHEADER);
+ bi.bmiHeader.biWidth=ww;bi.bmiHeader.biHeight=-wh;bi.bmiHeader.biPlanes=1;bi.bmiHeader.biBitCount=32;
+ void *bits{};HBITMAP bitmap=CreateDIBSection(dc,&bi,DIB_RGB_COLORS,&bits,nullptr,0);
+ if(!bitmap){DeleteDC(dc);return {};}
+ auto old=SelectObject(dc,bitmap);Pixels p;
+ if(PrintWindow(window,dc,2)) {
+  p.width=width;p.height=height;p.bgra.resize(size_t(width)*height*4);
+  for(int row=0;row<height;++row) memcpy(p.bgra.data()+size_t(row)*width*4,(BYTE*)bits+(size_t(y+row)*ww+x)*4,size_t(width)*4);
+ }
+ SelectObject(dc,old);DeleteObject(bitmap);DeleteDC(dc);return p;
+}
+}
+std::wstring cachedViberChat(HWND window) {
+ std::lock_guard lock(viberMutex);
+ return window==viberWindow && GetTickCount64()-viberTime<2000 ? viberKey : L"";
+}
+std::wstring recognizeViberChat(HWND window,const Settings &settings) {
+ std::wstring key;
+ try {
+  auto hr=CoInitializeEx(nullptr,COINIT_MULTITHREADED);
+  struct Apartment {bool owned;~Apartment(){if(owned)CoUninitialize();}} apartment{SUCCEEDED(hr)};
+  auto context=SetThreadDpiAwarenessContext(DPI_AWARENESS_CONTEXT_PER_MONITOR_AWARE_V2);
+  struct Dpi {DPI_AWARENESS_CONTEXT old;~Dpi(){if(old)SetThreadDpiAwarenessContext(old);}} dpi{context};
+  auto before=viberHeader(window);
+  if(!before.bgra.empty()) {
+   Settings options=settings;options.useTesseract=true;
+   Clip clip;clip.formats.push_back(imageDib(before));
+   auto title=recognizeText(clip,options);
+   while(!title.empty() && iswspace(title.back())) title.pop_back();
+   while(!title.empty() && iswspace(title.front())) title.erase(0,1);
+   auto after=viberHeader(window);
+   if(!title.empty() && title.size()<512 && title.find_first_of(L"\r\n")==title.npos && before.width==after.width && before.height==after.height && before.bgra==after.bgra)
+    key=L"viber\n"+title;
+  }
+ } catch(...) {}
+ {std::lock_guard lock(viberMutex);viberWindow=window;viberKey=key;viberTime=GetTickCount64();}
+ return key;
+}
+}