fix flameshot-org#702 add OCR with tesseract

data/graphics.qrc

@@ -96,8 +96,10 @@
         <file>img/material/white/move_up.svg</file>
         <file>img/material/white/delete.svg</file>
         <file>img/material/black/apps.svg</file>
+        <file>img/material/black/ocr.svg</file>
         <file>img/material/black/image.svg</file>
         <file>img/material/white/apps.svg</file>
+        <file>img/material/white/ocr.svg</file>
         <file>img/material/white/image.svg</file>
     </qresource>
 </RCC>

src/config/generalconf.cpp

@@ -56,6 +56,7 @@ GeneralConf::GeneralConf(QWidget* parent)
     initAntialiasingPinZoom();
     initUploadHistoryMax();
     initUndoLimit();
+    initTesseractPath();
     initUploadClientSecret();
     initPredefinedColorPaletteLarge();
     initShowSelectionGeometry();
@@ -545,6 +546,32 @@ void GeneralConf::initUploadHistoryMax()
     vboxLayout->addWidget(m_uploadHistoryMax);
 }
 
+void GeneralConf::initTesseractPath()
+{
+    auto* box = new QGroupBox(tr("Tesseract path (for OCR)"));
+    box->setFlat(true);
+    m_layout->addWidget(box);
+
+    auto* vboxLayout = new QVBoxLayout();
+    box->setLayout(vboxLayout);
+
+    m_tesseractPath = new QLineEdit(this);
+    QString foreground = this->palette().windowText().color().name();
+    m_tesseractPath->setStyleSheet(
+      QStringLiteral("color: %1").arg(foreground));
+    m_tesseractPath->setText(ConfigHandler().tesseractBinPath());
+    connect(m_tesseractPath,
+            SIGNAL(editingFinished()),
+            this,
+            SLOT(tesseractPathEdited()));
+    vboxLayout->addWidget(m_tesseractPath);
+}
+
+void GeneralConf::tesseractPathEdited()
+{
+    ConfigHandler().setTesseractBinPath(m_tesseractPath->text());
+}
+
 void GeneralConf::initUploadClientSecret()
 {
     auto* box = new QGroupBox(tr("Imgur Application Client ID"));

src/config/generalconf.h

@@ -52,6 +52,7 @@ private slots:
     void exportFileConfiguration();
     void resetConfiguration();
     void togglePathFixed();
+    void tesseractPathEdited();
     void uploadClientKeyEdited();
     void useJpgForClipboardChanged(bool checked);
     void setSaveAsFileExtension(QString extension);
@@ -87,6 +88,7 @@ private slots:
     void initUploadWithoutConfirmation();
     void initUseJpgForClipboard();
     void initUploadHistoryMax();
+    void initTesseractPath();
     void initUploadClientSecret();
     void initSaveLastRegion();
     void initShowSelectionGeometry();
@@ -118,6 +120,7 @@ private slots:
     QPushButton* m_resetButton;
     QCheckBox* m_saveAfterCopy;
     QLineEdit* m_savePath;
+    QLineEdit* m_tesseractPath;
     QLineEdit* m_uploadClientKey;
     QPushButton* m_changeSaveButton;
     QCheckBox* m_screenshotPathFixedCheck;

src/core/capturerequest.h

@@ -27,6 +27,7 @@ class CaptureRequest
         PIN = 16,
         UPLOAD = 32,
         ACCEPT_ON_SELECT = 64,
+        OCR = 128,
     };
 
     CaptureRequest(CaptureMode mode,

src/core/flameshot.cpp

@@ -26,6 +26,7 @@
 #include <QDebug>
 #include <QDesktopWidget>
 #include <QFile>
+#include <QProcess>
 #include <QMessageBox>
 #include <QThread>
 #include <QTimer>
@@ -349,6 +350,29 @@ void Flameshot::exportCapture(QPixmap capture,
           << selection.x() << "+" << selection.y() << "\n";
     }
 
+    if (tasks & CR::OCR) {
+        QProcess *tesseractProcess = new QProcess();
+
+        tesseractProcess->setProcessChannelMode(QProcess::MergedChannels);
+        tesseractProcess->start(ConfigHandler().tesseractBinPath(), QStringList() << "stdin" << "stdout");
+
+        capture.save(tesseractProcess, "PNG");
+        tesseractProcess->closeWriteChannel();
+
+        AbstractLogger::info() << QObject::tr("Running OCR with tesseract");
+
+        if(!tesseractProcess->waitForFinished()){
+            QMessageBox messageBox;
+            QString errorMessage = tr("Error running tesseract OCR from ") + ConfigHandler().tesseractBinPath() + "\n" + tesseractProcess->errorString();
+            messageBox.critical(0, tr("Error"), errorMessage);
+            messageBox.setFixedSize(500,200);
+        }else{
+            QString stdout = tesseractProcess->readAllStandardOutput();
+            FlameshotDaemon::copyToClipboard(stdout);
+        }
+
+    }
+
     if (tasks & CR::PRINT_RAW) {
         QByteArray byteArray;
         QBuffer buffer(&byteArray);

src/tools/CMakeLists.txt

@@ -41,6 +41,7 @@ target_sources(
 target_sources(flameshot PRIVATE rectangle/rectangletool.h rectangle/rectangletool.cpp)
 target_sources(flameshot PRIVATE redo/redotool.h redo/redotool.cpp)
 target_sources(flameshot PRIVATE save/savetool.h save/savetool.cpp)
+target_sources(flameshot PRIVATE ocr/ocrtool.h ocr/ocrtool.cpp)
 target_sources(flameshot PRIVATE accept/accepttool.h accept/accepttool.cpp)
 target_sources(flameshot PRIVATE invert/inverttool.h invert/inverttool.cpp)
 target_sources(flameshot PRIVATE selection/selectiontool.h selection/selectiontool.cpp)

src/tools/capturetool.h

@@ -48,6 +48,7 @@ class CaptureTool : public QObject
         TYPE_SIZEDECREASE = 21,
         TYPE_INVERT = 22,
         TYPE_ACCEPT = 23,
+        TYPE_OCR = 24,
     };
     Q_ENUM(Type);
 

src/tools/ocr/ocrtool.cpp

@@ -0,0 +1,55 @@
+// SPDX-License-Identifier: GPL-3.0-or-later
+// SPDX-FileCopyrightText: 2017-2019 Alejandro Sirgo Rica & Contributors
+
+#include "ocrtool.h"
+#include "src/utils/screenshotsaver.h"
+#include <QApplication>
+#include <QPainter>
+#include <QStyle>
+#if defined(Q_OS_MACOS)
+#include "src/widgets/capture/capturewidget.h"
+#include <QWidget>
+#endif
+
+OcrTool::OcrTool(QObject* parent)
+  : AbstractActionTool(parent)
+{}
+
+bool OcrTool::closeOnButtonPressed() const
+{
+    return true;
+}
+
+QIcon OcrTool::icon(const QColor& background, bool inEditor) const
+{
+    Q_UNUSED(inEditor)
+    return QIcon(iconPath(background) + "ocr.svg");
+}
+
+QString OcrTool::name() const
+{
+    return tr("OCR");
+}
+
+CaptureTool::Type OcrTool::type() const
+{
+    return CaptureTool::TYPE_OCR;
+}
+
+QString OcrTool::description() const
+{
+    return tr("OCR the capture");
+}
+
+CaptureTool* OcrTool::copy(QObject* parent)
+{
+    return new OcrTool(parent);
+}
+
+void OcrTool::pressed(CaptureContext& context)
+{
+    emit requestAction(REQ_CLEAR_SELECTION);
+    emit requestAction(REQ_CAPTURE_DONE_OK);
+    context.request.addTask(CaptureRequest::OCR);
+    emit requestAction(REQ_CLOSE_GUI);
+}

src/tools/ocr/ocrtool.h

@@ -0,0 +1,27 @@
+// SPDX-License-Identifier: GPL-3.0-or-later
+// SPDX-FileCopyrightText: 2017-2019 Alejandro Sirgo Rica & Contributors
+
+#pragma once
+
+#include "abstractactiontool.h"
+
+class OcrTool : public AbstractActionTool
+{
+    Q_OBJECT
+public:
+    explicit OcrTool(QObject* parent = nullptr);
+
+    bool closeOnButtonPressed() const override;
+
+    QIcon icon(const QColor& background, bool inEditor) const override;
+    QString name() const override;
+    QString description() const override;
+
+    CaptureTool* copy(QObject* parent = nullptr) override;
+
+protected:
+    CaptureTool::Type type() const override;
+
+public slots:
+    void pressed(CaptureContext& context) override;
+};

src/tools/toolfactory.cpp

@@ -14,6 +14,7 @@
 #include "line/linetool.h"
 #include "marker/markertool.h"
 #include "move/movetool.h"
+#include "ocr/ocrtool.h"
 #include "pencil/penciltool.h"
 #include "pin/pintool.h"
 #include "pixelate/pixelatetool.h"
@@ -61,6 +62,7 @@ CaptureTool* ToolFactory::CreateTool(CaptureTool::Type t, QObject* parent)
         if_TYPE_return_TOOL(TYPE_SIZEINCREASE, SizeIncreaseTool);
         if_TYPE_return_TOOL(TYPE_SIZEDECREASE, SizeDecreaseTool);
         if_TYPE_return_TOOL(TYPE_INVERT, InvertTool);
+        if_TYPE_return_TOOL(TYPE_OCR, OcrTool);
         if_TYPE_return_TOOL(TYPE_ACCEPT, AcceptTool);
         default:
             return nullptr;

src/utils/confighandler.cpp

@@ -123,6 +123,7 @@ static QMap<class QString, QSharedPointer<ValueHandler>>
     // drawFontSize, remember to update ConfigHandler::toolSize
     OPTION("copyOnDoubleClick"           ,Bool               ( false         )),
     OPTION("uploadClientSecret"          ,String             ( "313baf0c7b4d3ff"            )),
+    OPTION("tesseractBinPath"            ,String             ( "tesseract"   )),
     OPTION("showSelectionGeometry"  , BoundedInt               (0,5,4)),
     OPTION("showSelectionGeometryHideTime", LowerBoundedInt       (0, 3000))
 };
@@ -142,6 +143,7 @@ static QMap<QString, QSharedPointer<KeySequence>> recognizedShortcuts = {
     SHORTCUT("TYPE_SAVE"                ,   "Ctrl+S"                ),
     SHORTCUT("TYPE_ACCEPT"              ,   "Return"                ),
     SHORTCUT("TYPE_EXIT"                ,   "Ctrl+Q"                ),
+    SHORTCUT("TYPE_OCR"                 ,                           ),
     SHORTCUT("TYPE_IMAGEUPLOADER"       ,                           ),
 #if !defined(Q_OS_MACOS)
     SHORTCUT("TYPE_OPEN_APP"            ,   "Ctrl+O"                ),

src/utils/confighandler.h

@@ -124,6 +124,7 @@ class ConfigHandler : public QObject
     CONFIG_GETTER_SETTER(squareMagnifier, setSquareMagnifier, bool)
     CONFIG_GETTER_SETTER(copyOnDoubleClick, setCopyOnDoubleClick, bool)
     CONFIG_GETTER_SETTER(uploadClientSecret, setUploadClientSecret, QString)
+    CONFIG_GETTER_SETTER(tesseractBinPath, setTesseractBinPath, QString)
     CONFIG_GETTER_SETTER(saveLastRegion, setSaveLastRegion, bool)
     CONFIG_GETTER_SETTER(showSelectionGeometry, setShowSelectionGeometry, int)
     // SPECIAL CASES

src/widgets/capture/capturetoolbutton.cpp

@@ -157,6 +157,7 @@ static std::map<CaptureTool::Type, int> buttonTypeOrder
 
       { CaptureTool::TYPE_SIZEINCREASE, 22 },
       { CaptureTool::TYPE_SIZEDECREASE, 23 },
+      { CaptureTool::TYPE_OCR, 24 },
 };
 
 int CaptureToolButton::getPriorityByButton(CaptureTool::Type b)
@@ -175,7 +176,7 @@ QList<CaptureTool::Type> CaptureToolButton::iterableButtonTypes = {
     CaptureTool::TYPE_MOVESELECTION, CaptureTool::TYPE_UNDO,
     CaptureTool::TYPE_REDO,          CaptureTool::TYPE_COPY,
     CaptureTool::TYPE_SAVE,          CaptureTool::TYPE_EXIT,
-    CaptureTool::TYPE_IMAGEUPLOADER,
+    CaptureTool::TYPE_IMAGEUPLOADER, CaptureTool::TYPE_OCR,
 #if !defined(Q_OS_MACOS)
     CaptureTool::TYPE_OPEN_APP,
 #endif