/**************************************************************************** ** $Id: qinputcontext.cpp,v 1.6 2004/06/22 06:47:30 daisuke Exp $ ** ** Implementation of TQInputContext class ** ** Copyright (C) 2000-2003 Trolltech AS. All rights reserved. ** ** This file is part of the kernel module of the TQt GUI Toolkit. ** ** This file may be distributed under the terms of the Q Public License ** as defined by Trolltech AS of Norway and appearing in the file ** LICENSE.TQPL included in the packaging of this file. ** ** This file may be distributed and/or modified under the terms of the ** GNU General Public License version 2 as published by the Free Software ** Foundation and appearing in the file LICENSE.GPL included in the ** packaging of this file. ** ** Licensees holding valid TQt Enterprise Edition or TQt Professional Edition ** licenses for Unix/X11 may use this file in accordance with the TQt Commercial ** License Agreement provided with the Software. ** ** This file is provided AS IS with NO WARRANTY OF ANY KIND, INCLUDING THE ** WARRANTY OF DESIGN, MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. ** ** See http://www.trolltech.com/pricing.html or email sales@trolltech.com for ** information about TQt Commercial License Agreements. ** See http://www.trolltech.com/qpl/ for TQPL licensing information. ** See http://www.trolltech.com/gpl/ for GPL licensing information. ** ** Contact info@trolltech.com if any conditions of this licensing are ** not clear to you. ** **********************************************************************/ //#define TQT_NO_IM_PREEDIT_RELOCATION #include "tqinputcontext.h" #ifndef TQT_NO_IM #include "tqplatformdefs.h" #include "tqapplication.h" #include "tqwidget.h" #include "tqpopupmenu.h" #include #include class TQInputContextPrivate { public: TQInputContextPrivate() : holderWidget( 0 ), composingWidget( 0 ), hasFocus( FALSE ), isComposing( FALSE ) #if !defined(TQT_NO_IM_PREEDIT_RELOCATION) , preeditString( TQString() ), cursorPosition( -1 ), selLength ( 0 ) #endif {} TQWidget *holderWidget; // widget to which TQInputContext instance belongs. TQWidget *composingWidget; bool hasFocus; bool isComposing; void updateComposingState( const TQString &text, int newCursorPosition, int newSelLength ) { #if !defined(TQT_NO_IM_PREEDIT_RELOCATION) preeditString = text; cursorPosition = newCursorPosition; selLength = newSelLength; #endif } void resetComposingState() { isComposing = FALSE; #if !defined(TQT_NO_IM_PREEDIT_RELOCATION) preeditString = TQString(); cursorPosition = -1; selLength = 0; #endif } #if !defined(TQT_NO_IM_PREEDIT_RELOCATION) TQString preeditString; int cursorPosition; int selLength; #endif }; // UPDATED COMMENT RETQUIRED -- 2004-07-08 YamaKen /*! \class TQInputContext qinputcontext.h \brief The TQInputContext class abstracts the input method dependent data and composing state. \ingroup i18n An input method is responsible to input complex text that cannot be inputted via simple keymap. It converts a sequence of input events (typically key events) into a text string through the input method specific converting process. The class of the processes are widely ranging from simple finite state machine to complex text translator that pools a whole paragraph of a text with text editing capability to perform grammar and semantic analysis. To abstract such different input method specific intermediate information, TQt offers the TQInputContext as base class. The concept is well known as 'input context' in the input method domain. an input context is created for a text widget in response to a demand. It is ensured that an input context is prepared for an input method before input to a text widget. Multiple input contexts that is belonging to a single input method may concurrently coexist. Suppose multi-window text editor. Each text widget of window A and B holds different TQInputContext instance which contains different state information such as partially composed text. \section1 Groups of functions: \table \header \i Context \i Functions \row \i Receiving information \i x11FilterEvent(), filterEvent(), setMicroFocus(), mouseHandler() \row \i Sending back composed text \i sendIMEvent(), \row \i State change notification \i setFocus(), unsetFocus(), reset() \row \i Context information \i identifierName(), language(), font(), isComposing(), \endtable \section1 Sharing input context between text widgets Any input context can be shared between several text widgets to reduce resource consumption. In ideal case, each text widgets should be allocated dedicated input context. But some complex input contexts require slightly heavy resource such as 100 kilobytes of memory. It prevents quite many text widgets from being used concurrently. To resolve such problem, we can share an input context. There is one 'input context holder widget' per text widgets that shares identical input context. In this model, the holder widget owns the shared input context. Other text widgets access the input context via TQApplication::locateICHolderWidget(). But the access convention is transparently hidden into TQWidget, so developers are not required to aware of it. What developer should know is only the mapping function TQApplication::locateICHolderWidget(). It accepts a widget as argument and returns its holder widget. Default implementation returns the top-level widget of the widget as reasonable assumption. But some applications should reimplement the function to fit application specific usability. See TQApplication::locateICHolderWidget() for further information. \section1 Preedit preservation As described above, input contexts have wide variety of amount of the state information in accordance with belonging input method. It is ranging from 2-3 keystrokes of sequence in deterministic input methods to hundreds of keystrokes with semantic text refinement in complex input methods such as ordinary Japanese input method. The difference requires the different reset policies in losing input focus. The former simple input method case, users will prefer resetting the context to back to the neutral state when something happened. Suppose a web browsing. The user scroll the page by scrollbar after he or she has typed a half of the valid key sequence into a text widget. In the case, the input context should be reset in losing focus when he or she has dragged the scrollbar. He or she will be confused if the input context is still preserved until focused back to the text widget because he or she will restart typing with first key of the sequence as a habitual operation. On the other hand, we should choose completely different policy for the latter complex input method case. Suppose same situation as above but he or she is using a complex input method. In the case, he or she will be angry if the input context has been lost when he or she has dragged the scrollbar because the input context contained a valuably composed text made up by considerable input cost. So we should not reset the input context in the case. And the input context should be preserved until focused back to the text widget. This behavior is named as 'preedit preservation'. The two policies can be switched by calling or not calling reset() in unsetFocus(). Default implementation of unsetFocus() calls reset() to fit the simple input methods. The implementation is expressed as 'preedit preservation is disabled'. \section1 Preedit relocation Although the most case of the preedit preservation problem for complex input methods is resolved as described above, there is a special case. Suppose the case that matches all of the following conditions. \list \i a input focus has been moved from a text widget to another text widget directly \i the input context is shared between the two text widgets \i preedit preservation is enabled for the input context \endlist In the case, there are the following two requirements that contradicts each other. The input context sharing causes it. \list \i the input context has to be reset to prepare to input to the newly focused text widget \i the input context has to be preserved until focused back to the previous text widget \endlist A intrinsic feature named 'preedit relocation' is available to compromise the requirements. If the feature is enabled for the input context, it is simply moved to the new text widget with the preedit string. The user continues the input on the new text widget, or relocate it to another text widget. The preedit of previous text widget is automatically cleared to back to the neutral state of the widget. This strange behavior is just a compromise. As described in previous section, complex input method user should not be exposed to the risk losing the input context because it contains valuable long text made up with considerable input cost. The user will immediately focus back to the previous text widget to continue the input in the correct text widget if the preedit relocation occurred. The feature is mainly existing as safety. The feature properly works even if the focus is moved as following. Input method developers are not required to be aware of the relocation protocol since TQInputContext transparently handles it. a text widget -> a non-text widget -> another text widget To enable the preedit relocation feature, the input context class have to reimplement isPreeditRelocationEnabled() as returns TRUE. The implementation requires that the preedit preservation is also enabled since preedit relocation is a special case of the preedit preservation. If the preedit relocation is disabled, the input context is simply reset in the relocation case. \section1 Input context instanciation \section1 Input method switching \section1 Text widget implementor's guide Add following code fragment into createPopupMenu() to add input method dependent submenus. \code #ifndef TQT_NO_IM TQInputContext *qic = getInputContext(); if ( qic ) qic->addMenusTo( popup ); #endif \endcode \sa TQInputContextPlugin, TQInputContextFactory, TQApplication::locateICHolderWidget(), TQApplication::defaultInputMethod() */ /*! Constructs an input context. holderWidget is set immediately after this constructor has been returned on the X11 platform. */ TQInputContext::TQInputContext( TQObject *tqparent ) : TQObject( tqparent ) { d = new TQInputContextPrivate; } /*! Destroys the input context. */ TQInputContext::~TQInputContext() { delete d; } #if defined(TQ_WS_X11) /*! \internal Returns the owner of this input context. Ordinary input methods should not call this function directly to keep platform independence and flexible configuration possibility. The return value may differ from tqfocusWidget() if the input context is shared between several text widgets. \sa setHolderWidget(), tqfocusWidget() */ TQWidget *TQInputContext::holderWidget() const { return d->holderWidget; } /*! \internal Sets the owner of this input context. Ordinary input methods must not call this function directly. \sa holderWidget() */ void TQInputContext::setHolderWidget( TQWidget *w ) { d->holderWidget = w; } /*! \internal Returns the widget that has an input focus for this input context. Ordinary input methods should not call this function directly to keep platform independence and flexible configuration possibility. The return value may differ from holderWidget() if the input context is shared between several text widgets. \sa setFocusWidget(), holderWidget() */ TQWidget *TQInputContext::tqfocusWidget() const { return d->hasFocus ? d->composingWidget : 0; } /*! \internal Sets the widget that has an input focus for this input context. Ordinary input methods must not call this function directly. \sa tqfocusWidget() */ void TQInputContext::setFocusWidget( TQWidget *w ) { if ( w ) { bool isFocusingBack = ( w == d->composingWidget ); bool isPreeditRelocation = ( ! isFocusingBack && isComposing() && d->composingWidget ); // invoke sendIMEventInternal() rather than sendIMEvent() to // avoid altering the composing state if ( isPreeditRelocation == TRUE ) { // clear preedit of previously focused text // widget. preserved preedit may be exist even if // isPreeditRelocationEnabled() == FALSE. sendIMEventInternal( TQEvent::IMEnd ); } d->composingWidget = w; // changes recipient of TQIMEvent if ( isPreeditRelocation == TRUE ) { #if !defined(TQT_NO_IM_PREEDIT_RELOCATION) if ( isPreeditRelocationEnabled() ) { // copy preedit state to the widget that gaining focus sendIMEventInternal( TQEvent::IMStart ); sendIMEventInternal( TQEvent::IMCompose, d->preeditString, d->cursorPosition, d->selLength ); } else #endif { // reset input context when the shared context has // focused on another text widget reset(); } } } d->hasFocus = w ? TRUE : FALSE; } /*! \internal This function is called from TQWidget to keep input state consistency. Ordinary input method must not call this function directly. */ void TQInputContext::releaseComposingWidget( TQWidget *w ) { if ( d->composingWidget == w ) { d->composingWidget = 0; d->hasFocus = FALSE; } } #endif // TQ_WS_X11 /*! \internal This function can be reimplemented in a subclass as returning TRUE if you want making your input method enable the preedit relocation. See the description for preedit relocation of TQInputContext. /sa TQInputContext */ bool TQInputContext::isPreeditRelocationEnabled() { return FALSE; } /*! This function indicates whether IMStart event had been sent to the text widget. It is ensured that an input context can send IMCompose or IMEnd event safely if this function returned TRUE. The state is automatically being tracked through sendIMEvent(). \sa sendIMEvent() */ bool TQInputContext::isComposing() const { return d->isComposing; } /*! This function can be reimplemented in a subclass to filter input events. Return TRUE if the \a event has been consumed. Otherwise, the unfiltered \a event will be forwarded to widgets as ordinary way. Although the input events have accept() and ignore() methods, leave it untouched. \a event is currently restricted to TQKeyEvent. But some input method related events such as TQWheelEvent or TQTabletEvent may be added in future. The filtering opportunity is always given to the input context as soon as possible. It has to be taken place before any other key event consumers such as eventfilters and accelerators because some input methods require quite various key combination and sequences. It often conflicts with accelerators and so on, so we must give the input context the filtering opportunity first to ensure all input methods work properly regardless of application design. Ordinary input methods require discrete key events to work properly, so TQt's key compression is always disabled for any input contexts. \sa TQKeyEvent, x11FilterEvent() */ bool TQInputContext::filterEvent( const TQEvent *event ) { return FALSE; } /*! \fn void TQInputContext::deletionRequested() Emit this signal when a fatal error has been caused in the input context. The input context will be deleted by the owner which is usually the holder widget. */ /*! \fn void TQInputContext::imEventGenerated( TQObject *receiver, TQIMEvent *e ) \internal This signal is emitted when the user has sent a TQIMEvent through sendIMEvent(). Ordinary input methods should not emit this signal directly. \a receiver is a platform dependent destination of the \a e. \sa TQIMEvent, sendIMEvent(), sendIMEventInternal(), */ /*! \internal Sends a TQIMEvent to the client via imEventGenerated() signal. Ordinary input method should not call this function directly. \sa TQIMEvent, TQIMComposeEvent, sendIMEvent(), imEventGenerated() */ void TQInputContext::sendIMEventInternal( TQEvent::Type type, const TQString &text, int cursorPosition, int selLength ) { TQObject *receiver = 0; TQIMEvent *event = 0; #if defined(TQ_WS_X11) receiver = TQT_TQOBJECT(d->composingWidget); #elif defined(TQ_WS_TQWS) // just a placeholder #endif if ( ! receiver ) return; if ( type == TQEvent::IMStart ) { qDebug( "sending IMStart with %d chars to %p", text.length(), receiver ); event = new TQIMEvent( type, text, cursorPosition ); } else if ( type == TQEvent::IMEnd ) { qDebug( "sending IMEnd with %d chars to %p, text=%s", text.length(), receiver, (const char*)text.local8Bit() ); event = new TQIMEvent( type, text, cursorPosition ); } else if ( type == TQEvent::IMCompose ) { qDebug( "sending IMCompose to %p with %d chars, cpos=%d, sellen=%d, text=%s", receiver, text.length(), cursorPosition, selLength, (const char*)text.local8Bit() ); event = new TQIMComposeEvent( type, text, cursorPosition, selLength ); } if ( event ) emit imEventGenerated( receiver, event ); } /*! Call this function to send TQIMEvent to the text widget. This function constructs a TQIMEvent based on the arguments and send it to the appropriate widget. Ordinary input method should not reimplement this function. \a type is either \c TQEvent::IMStart or \c TQEvent::IMCompose or \c TQEvent::IMEnd. You have to send a \c TQEvent::IMStart to start composing, then send several \c TQEvent::IMCompose to update the preedit of the widget, and finalize the composition with sending \c TQEvent::IMEnd. \c TQEvent::IMStart should always be sent without arguments as: \code sendIMEvent( TQEvent::IMStart ) \endcode And \c TQEvent::IMCompose can be sent without cursor: \code sendIMEvent( TQEvent::IMCompose, TQString( "a text" ) ) \endcode Or optionally with cursor with \a cursorPosition: \code sendIMEvent( TQEvent::IMCompose, TQString( "a text with cursor" ), 12 ) \endcode Note that \a cursorPosition also specifies microfocus position. Or optionally with selection text: \code sendIMEvent( TQEvent::IMCompose, TQString( "a text with selection" ), 12, 9 ) \endcode \a cursorPosition and \a selLength must be within the \a text. The \a cursorPosition also specifies microfocus position in the case: \c TQEvent::IMEnd can be sent without arguments to terminate the composition with null string: \code sendIMEvent( TQEvent::IMEnd ) \endcode Or optionally accepts \a text to commit a string: \code sendIMEvent( TQEvent::IMEnd, TQString( "a text" ) ) \endcode \sa TQIMEvent, TQIMComposeEvent, setMicroFocus() */ void TQInputContext::sendIMEvent( TQEvent::Type type, const TQString &text, int cursorPosition, int selLength ) { #if defined(TQ_WS_X11) if ( !tqfocusWidget() ) return; #endif if ( type == TQEvent::IMStart ) { sendIMEventInternal( type, text, cursorPosition, selLength ); d->isComposing = TRUE; } else if ( type == TQEvent::IMEnd ) { d->resetComposingState(); sendIMEventInternal( type, text, cursorPosition, selLength ); } else if ( type == TQEvent::IMCompose ) { d->updateComposingState( text, cursorPosition, selLength ); sendIMEventInternal( type, text, cursorPosition, selLength ); } } /*! This function can be reimplemented in a subclass to detect that the input context has been focused on. The input context will receive input events through x11FilterEvent() and filterEvent() after setFocus() until unsetFocus() has been called. an input context is ensured that setFocus() is called exactly once until unsetFocus() has been called even if preedit relocation has occurred. This means that an input focus will survive between several widgets that sharing the input context. On the X11 platform, tqfocusWidget is already set before this function has been called. \sa unsetFocus() */ void TQInputContext::setFocus() { } /*! This function can be reimplemented in a subclass to detect that the input context has lost the focus. an input context is ensured that unsetFocus() is not called during preedit relocation. This means that an input focus will survive between several widgets that sharing the input context. Default implementation that calls reset() is sufficient for simple input methods. You can override this function to alter the behavior. For example, most Japanese input contexts should not be reset on losing focus. The context sometimes contains a whole paragraph and has minutes of lifetime different to ephemeral one in other languages. The piled input context should be survived until focused again since Japanese user naturally expects so. On the X11 platform, tqfocusWidget is valid until this function has been returned. \sa setFocus() */ void TQInputContext::unsetFocus() { reset(); } /*! This function can be implemented in a subclass to handle microfocus changes. 'microfocus' stands for the input method focus point in the preedit (XIM "spot" point) for complex language input handling. It can be used to place auxiliary GUI widgets such as candidate selection window. \a x, \a y, \a w and \a h represents the position and size of the cursor in the preedit string. \a f is the font on the location of the cursor. */ void TQInputContext::setMicroFocus( int x, int y, int w, int h, TQFont *f ) { } /*! This function can be reimplemented in a subclass to handle mouse presses/releases/doubleclicks/moves within the preedit text. You can use the function to implement mouse-oriented user interface such as text selection or popup menu for candidate selection. The parameter \a x is the offset within the string that was sent with the IMCompose event. The alteration boundary of \a x is ensured as character boundary of preedit string accurately. \a type is either \c TQEvent::MouseButtonPress or \c TQEvent::MouseButtonRelease or \c TQEvent::MouseButtonDblClick or \c TQEvent::MouseButtonMove. Refer \a button and \a state to determine what operation has performed. The method interface is imported from TQWSInputMethod::mouseHandler() of TQt/Embedded 2.3.7 and extended for desktop system. */ void TQInputContext::mouseHandler( int x, TQEvent::Type type, TQt::ButtonState button, TQt::ButtonState state ) { // Default behavior for simple ephemeral input contexts. Some // complex input contexts should not be reset here. if ( type == TQEvent::MouseButtonPress || type == TQEvent::MouseButtonDblClick ) reset(); } /*! Returns the font of the current input widget */ TQFont TQInputContext::font() const { if ( !tqfocusWidget() ) return TQApplication::font(); //### absolutely last resort return tqfocusWidget()->font(); } /*! This function can be reimplemented in a subclass to reset the state of the input method. This function is called by several widgets to reset input state. For example, a text widget call this function before inserting a text to make widget ready to accept a text. Default implementation is sufficient for simple input method. You can override this function to reset external input method engines in complex input method. In the case, call TQInputContext::reset() to ensure proper termination of inputting. You must not send any TQIMEvent except empty IMEnd event using TQInputContext::reset() at reimplemented reset(). It will break input state consistency. */ void TQInputContext::reset() { if ( isComposing() ) sendIMEvent( TQEvent::IMEnd ); } /*! This function must be implemented in any subclasses to return the identifier name of the input method. Return value is the name to identify and specify input methods for the input method switching mechanism and so on. The name has to be consistent with TQInputContextPlugin::keys(). The name has to consist of ASCII characters only. There are two different names with different responsibility in the input method domain. This function returns one of them. Another name is called 'display name' that stands for the name for endusers appeared in a menu and so on. \sa TQInputContextPlugin::keys(), TQInputContextPlugin::displayName() */ TQString TQInputContext::identifierName() { return ""; } /*! This function must be implemented in any subclasses to return a language code (e.g. "zh_CN", "zh_TW", "zh_HK", "ja", "ko", ...) of the input context. If the input context can handle multiple languages, return the currently used one. The name has to be consistent with TQInputContextPlugin::language(). This information will be used by language tagging feature in TQIMEvent. It is required to distinguish unified han characters correctly. It enables proper font and character code handling. Suppose CJK-awared multilingual web browser (that automatically modifies fonts in CJK-mixed text) and XML editor (that automatically inserts lang attr). \sa TQInputContextPlugin::language() */ TQString TQInputContext::language() { return ""; } /*! This function can be reimplemented in a subclass to provide input method dependent popup menus. Return 0 if the menus are unnecessary. Ownership of the object and tqchildren are transferred to the caller, and the result must not be called setAutoDelete(). TQInputContextMenu::title is used for label text of the popup menu as submenu. \sa addMenusTo() */ TQPtrList *TQInputContext::menus() { return 0; } /*! Appends input method dependent submenus into \a popup. A separator is also inserted into \a popup if \a action is InsertSeparator. This is an utility function only for convenience in limited situation. This function is used by input context owner such as text widgets to add the submenus to its own context menu. If you want to insert the submenus in more flexible way, use TQInputContext::menus() manually. \a popup is not restricted to context menu of a text widget. For example, the owner may be a input method menu of TQtopia taskbar in TQt/Embedded platform. \sa menus(), TQInputContextMenu::Action */ void TQInputContext::addMenusTo( TQPopupMenu *popup, TQInputContextMenu::Action action ) { if ( ! popup ) return; TQPtrList *imMenus = menus(); if ( imMenus ) { if ( action == TQInputContextMenu::InsertSeparator ) popup->insertSeparator(); for ( TQPtrList::Iterator it = imMenus->begin(); it != imMenus->end(); ++it ) { TQInputContextMenu *imMenu = *it; popup->insertItem( imMenu->title, imMenu->popup ); } imMenus->clear(); delete imMenus; } } #endif //TQ_NO_IM