always define wxUSE_RICHEDIT[2] -- this is better than never defining them (closes...

[wxWidgets.git] / docs / doxygen / overviews / eventhandling.h

/////////////////////////////////////////////////////////////////////////////
// Name:        eventhandling.h
// Purpose:     topic overview
// Author:      wxWidgets team
// RCS-ID:      $Id$
// Licence:     wxWindows license
/////////////////////////////////////////////////////////////////////////////

/**

@page overview_eventhandling Event Handling

Classes: wxEvtHandler, wxWindow, wxEvent

@li @ref overview_eventhandling_introduction
@li @ref overview_eventhandling_eventtables
@li @ref overview_eventhandling_connect
@li @ref overview_eventhandling_processing
@li @ref overview_eventhandling_prog
@li @ref overview_eventhandling_pluggable
@li @ref overview_eventhandling_winid
@li @ref overview_eventhandling_custom
@li @ref overview_eventhandling_macros


<hr>


@section overview_eventhandling_introduction Introduction

There are two principal ways to handle events in wxWidgets. One of them uses
<em>event table</em> macros and allows to define the connection between events
and their handlers only statically, i.e. during program compilation. The other
one uses wxEvtHandler::Connect() call and can be used to connect, and
disconnect, the handlers dynamically, i.e. during run-time depending on some
conditions. It also allows directly connecting the events of one object to a
handler method in another object while the static event tables can only handle
events in the object where they are defined so using Connect() is more flexible
than using the event tables. On the other hand, event tables are more succinct
and centralize all event handlers connection in one place. You can either
choose a single approach which you find preferable or freely combine both
methods in your program in different classes or even in one and the same class,
although this is probably sufficiently confusing to be a bad idea.

But before you make this choice, let us discuss these two ways in some more
details: in the next section we provide a short introduction to handling the
events using the event tables, please see @ref overview_eventhandling_connect
for the discussion of Connect().

@section overview_eventhandling_eventtables Event Handling with Event Tables

To use an <em>event table</em> you must first decide in which class you wish to
handle the events. The only requirement imposed by wxWidgets is that this class
must derive from wxEvtHandler and so, considering that wxWindow derives from
it, any classes representing windows can handle events. Simple events such as
menu commands are usually processed at the level of a top-level window
containing the menu, so let's suppose that you need to handle some events in @c
MyFrame class deriving from wxFrame.

First thing to do is to define one or more <em>event handlers</em>. They
are just simple (non-virtual) methods of the class which take as a parameter a
reference to an object of wxEvent-derived class and have no return value (any
return information is passed via the argument, which is why it is non-const).
You also need to insert a macro

@code
DECLARE_EVENT_TABLE()
@endcode

somewhere in the class declaration. It doesn't matter where does it occur but
it's customary to put it at the end of it because the macro changes the access
type internally and so it's safest if there is nothing that follows it. So the
full class declaration might look like this:

@code
class MyFrame : public wxFrame
{
public:
    MyFrame(...) : wxFrame(...) { }

    ...

protected:
    int m_whatever;

private:
    // notice that as the event handlers normally are not called from outside
    // the class, they normally be private, in particular they don't need at
    // all to be public
    void OnExit(wxCommandEvent& event);
    void OnButton1(wxCommandEvent& event);
    void OnSize(wxSizeEvent& event);

    // it's common to call the event handlers OnSomething() but there is no
    // obligation to it, this one is an event handler too:
    void DoTest(wxCommandEvent& event);

    DECLARE_EVENT_TABLE()
};
@endcode

Next the event table must be defined and, as any definition, it must be placed
in an implementation file to tell. The event table tells wxWidgets how to map
events to member functions and in our example it could look like this:

@code
BEGIN_EVENT_TABLE(MyFrame, wxFrame)
    EVT_MENU(wxID_EXIT, MyFrame::OnExit)
    EVT_MENU(DO_TEST, MyFrame::DoTest)
    EVT_SIZE(MyFrame::OnSize)
    EVT_BUTTON(BUTTON1, MyFrame::OnButton1)
END_EVENT_TABLE()
@endcode

Notice that you must mention a method you want to use for the event handling in
the event table definition, just defining it in MyFrame class is @e not enough.

Let us now look at the details of this definition: the first line means that we
are defining the event table for MyFrame class and that its base class is
wxFrame, so events not processed by MyFrame will, by default, be handled to
wxFrame. The next four lines define connections of individual events to their
handlers: the first two of them map menu commands from the items with the
identifiers specified as the first macro parameter to two different member
functions. In the next one, @c EVT_SIZE means that any changes in the size of
the frame will result in calling OnSize() method. Note that this macro doesn't
need a window identifier, since normally you are only interested in the current
window's size events.

The EVT_BUTTON macro demonstrates that the originating event does not have to
come from the window class implementing the event table -- if the event source
is a button within a panel within a frame, this will still work, because event
tables are searched up through the hierarchy of windows for the command events
(but only command events, so you can't catch mouse move events in a child
control in the parent window in the same way because wxMouseEvent doesn't
derive from wxCommandEvent, see below for how you can do it). In this case, the
button's event table will be searched, then the parent panel's, then the
frame's.

Finally, you need to implement the event handlers. As mentioned before, all
event handlers take a wxEvent-derived argument whose exact class differs
according to the type of event and the class of the originating window. For
size events, wxSizeEvent is used. For menu commands and most control commands
(such as button presses), wxCommandEvent is used. And when controls get more
complicated, more specific wxCommandEvent-derived event classes providing
additional control-specific information can be used, such as wxTreeEvent for
events from wxTreeCtrl windows.

In the simplest possible case an event handler may not use the @c event
parameter at all, e.g.

@code
void MyFrame::OnExit(wxCommandEvent&)
{
    // when the user selects "Exit" from the menu we should close
    Close(true);
}
@endcode

In other cases you may need some information carried by the @c event argument,
as in:

@code
void MyFrame::OnSize(wxSizeEvent& event)
{
    wxSize size = event.GetSize();

    ... update the frame using the new size ...
}
@endcode

You will find the details about the event table macros and the corresponding
wxEvent-derived classes in the discussion of each control generating these
events.


@section overview_eventhandling_connect Dynamic Event Handling

As with the event tables, you need to decide in which class do you intend to
handle the events first and, also as before, this class must still derive from
wxEvtHandler (usually indirectly via wxWindow), see the declaration of MyFrame
in the previous section. However the similarities end here and both the syntax
and the possibilities of this way of handling events in this way are rather
different.

Let us start by looking at the syntax: the first obvious difference is that you
don't need to use neither @c DECLARE_EVENT_TABLE() nor @c BEGIN_EVENT_TABLE and 
associated macros any more. Instead, in any place in your code, but usually in
the code of the class defining the handlers itself (and definitely not in the
global scope as with the event tables), you should call its Connect() method
like this:

@code
MyFrame::MyFrame(...)
{
      Connect(wxID_EXIT, wxEVT_COMMAND_MENU_SELECTED,
                wxCommandEventHandler(MyFrame::OnExit));
}
@endcode

This class should be self-explanatory except for wxCommandEventHandler part:
this is a macro which ensures that the method is of correct type by using
static_cast in the same way as event table macros do it inside them.

Now let us describe the semantic differences:
<ul>
    <li>
        Event handlers can be connected at any moment, e.g. it's possible to do
        some initialization first and only connect the handlers if and when it
        succeeds. This can avoid the need to test that the object was properly
        initialized in the event handlers themselves: with Connect() they
        simply won't be called at all if it wasn't.
    </li>

    <li>
        As a slight extension of the above, the handlers can also be
        Disconnect()-ed at any time. And maybe later reconnected again. Of
        course, it's also possible to emulate this behaviour with the classic
        static (i.e. connected via event tables) handlers by using an internal
        flag indicating whether the handler is currently enabled and returning
        from it if it isn't, but using dynamically connected handlers requires
        less code and is also usually more clear.
    </li>

    <li>
        Also notice that you must derive a class inherited from, say,
        wxTextCtrl even if you don't want to modify the control behaviour at
        all but just want to handle some of its events. This is especially
        inconvenient when the control is loaded from the XRC. Connecting the
        event handler dynamically bypasses the need for this unwanted
        sub-classing.
    </li>

    <li>
        Last but very, very far from least is the possibility to connect an
        event of some object to a method of another object. This is impossible
        to do with event tables because there is no possibility to specify the
        object to dispatch the event to so it necessarily needs to be sent to
        the same object which generated the event. Not so with Connect() which
        has an optional @c eventSink parameter which can be used to specify the
        object which will handle the event. Of course, in this case the method
        being connected must belong to the class which is the type of the
        @c eventSink object! To give a quick example, people often want to catch
        mouse movement events happening when the mouse is in one of the frame
        children in the frame itself. Doing it in a naive way doesn't work:
        <ul>
            <li>
                A @c EVT_LEAVE_WINDOW(MyFrame::OnMouseLeave) line in the frame
                event table has no effect as mouse move (including entering and
                leaving) events are not propagated upwards to the parent window
                (at least not by default).
            </li>

            <li>
                Putting the same line in a child event table will crash during
                run-time because the MyFrame method will be called on a wrong
                object -- it's easy to convince oneself that the only object
                which can be used here is the pointer to the child, as
                wxWidgets has nothing else. But calling a frame method with the
                child window pointer instead of the pointer to the frame is, of
                course, disastrous.
            </li>
        </ul>

        However writing
        @code
            MyFrame::MyFrame(...)
            {
              m_child->Connect(wxID_ANY, wxEVT_LEAVE_WINDOW,
                               wxMouseEventHandler(MyFrame::OnMouseLeave),
                               NULL,  // unused extra data parameter
                               this); // this indicates the object to connect to
            }
        @endcode
        will work exactly as expected. Note that you can get the object which
        generated the event -- and which is not the same as the frame -- via
        wxEvent::GetEventObject() method of @c event argument passed to the
        event handler.
    <li>
</ul>

To summarize, using Connect() requires slightly more typing but is much more
flexible than using static event tables so don't hesitate to use it when you
need this extra power. On the other hand, event tables are still perfectly fine
in simple situations where this extra flexibility is not needed.


@section overview_eventhandling_processing How Events are Processed

When an event is received from the windowing system, wxWidgets calls
wxEvtHandler::ProcessEvent on the first
event handler object belonging to the window generating the event.

It may be noted that wxWidgets' event processing system implements something
very close to virtual methods in normal C++, i.e. it is possible to alter
the behaviour of a class by overriding its event handling functions. In
many cases this works even for changing the behaviour of native controls.

For example it is possible to filter out a number of key events sent by the
system to a native text control by overriding wxTextCtrl and defining a
handler for key events using EVT_KEY_DOWN. This would indeed prevent
any key events from being sent to the native control - which might not be
what is desired. In this case the event handler function has to call Skip()
so as to indicate that the search for the event handler should continue.

To summarize, instead of explicitly calling the base class version as you
would have done with C++ virtual functions (i.e. @e wxTextCtrl::OnChar()),
you should instead call wxEvent::Skip.

In practice, this would look like this if the derived text control only
accepts 'a' to 'z' and 'A' to 'Z':

@code
void MyTextCtrl::OnChar(wxKeyEvent& event)
{
    if ( isalpha( event.KeyCode() ) )
    {
        // key code is within legal range. we call event.Skip() so the
        // event can be processed either in the base wxWidgets class
        // or the native control.

        event.Skip();
    }
    else
    {
        // illegal key hit. we don't call event.Skip() so the
        // event is not processed anywhere else.

        wxBell();
    }
}
@endcode

The normal order of event table searching by ProcessEvent is as follows:

@li If the object is disabled (via a call to wxEvtHandler::SetEvtHandlerEnabled)
    the function skips to step (6).
@li If the object is a wxWindow, @b ProcessEvent is recursively called on the window's
    wxValidator. If this returns @true, the function exits.
@li @b SearchEventTable is called for this event handler. If this fails, the base
    class table is tried, and so on until no more tables exist or an appropriate
    function was found, in which case the function exits.
@li The search is applied down the entire chain of event handlers (usually the chain has
    a length of one). If this succeeds, the function exits.
@li If the object is a wxWindow and the event is set to set to propagate (in the library only
    wxCommandEvent based events are set to propagate), @b ProcessEvent is recursively applied
    to the parent window's event handler. If this returns @true, the function exits.
@li Finally, @b ProcessEvent is called on the wxApp object.

<b>Pay close attention to Step 5</b>.  People often overlook or get
confused by this powerful feature of the wxWidgets event processing
system.  To put it a different way, events set to propagate
(see wxEvent::ShouldPropagate)
(most likely derived either directly or indirectly from wxCommandEvent)
will travel up the containment hierarchy from child to parent until the
maximal propagation level is reached or an event handler is found that
doesn't call @c event.Skip().

Finally, there is another additional complication (which, in fact, simplifies
life of wxWidgets programmers significantly): when propagating the command
events upwards to the parent window, the event propagation stops when it
reaches the parent dialog, if any. This means that you don't risk to get
unexpected events from the dialog controls (which might be left unprocessed by
the dialog itself because it doesn't care about them) when a modal dialog is
popped up. The events do propagate beyond the frames, however. The rationale
for this choice is that there are only a few frames in a typical application
and their parent-child relation are well understood by the programmer while it
may be very difficult, if not impossible, to track down all the dialogs which
may be popped up in a complex program (remember that some are created
automatically by wxWidgets). If you need to specify a different behaviour for
some reason, you can use wxWindow::SetExtraStyle(wxWS_EX_BLOCK_EVENTS)
explicitly to prevent the events from being propagated beyond the given window
or unset this flag for the dialogs which have it on by default.

Typically events that deal with a window as a window (size, motion,
paint, mouse, keyboard, etc.) are sent only to the window.  Events
that have a higher level of meaning and/or are generated by the window
itself, (button click, menu select, tree expand, etc.) are command
events and are sent up to the parent to see if it is interested in the event.

Note that your application may wish to override ProcessEvent to redirect processing of
events. This is done in the document/view framework, for example, to allow event handlers
to be defined in the document or view. To test for command events (which will probably
be the only events you wish to redirect), you may use wxEvent::IsCommandEvent for efficiency,
instead of using the slower run-time type system.

As mentioned above, only command events are recursively applied to the parents event
handler in the library itself. As this quite often causes confusion for users,
here is a list of system events which will NOT get sent to the parent's event handler:

@li wxEvent: The event base class
@li wxActivateEvent: A window or application activation event
@li wxCloseEvent: A close window or end session event
@li wxEraseEvent: An erase background event
@li wxFocusEvent: A window focus event
@li wxKeyEvent: A keypress event
@li wxIdleEvent: An idle event
@li wxInitDialogEvent: A dialog initialisation event
@li wxJoystickEvent: A joystick event
@li wxMenuEvent: A menu event
@li wxMouseEvent: A mouse event
@li wxMoveEvent: A move event
@li wxPaintEvent: A paint event
@li wxQueryLayoutInfoEvent: Used to query layout information
@li wxSetCursorEvent: Used for special cursor processing based on current mouse position
@li wxSizeEvent: A size event
@li wxScrollWinEvent: A scroll event sent by a scrolled window (not a scroll bar)
@li wxSysColourChangedEvent: A system colour change event

In some cases, it might be desired by the programmer to get a certain number
of system events in a parent window, for example all key events sent to, but not
used by, the native controls in a dialog. In this case, a special event handler
will have to be written that will override ProcessEvent() in order to pass
all events (or any selection of them) to the parent window.


@section overview_eventhandling_prog User Generated Events vs Programmatically Generated Events

While generically wxEvents can be generated both by user
actions (e.g. resize of a wxWindow) and by calls to functions
(e.g. wxWindow::SetSize), wxWidgets controls normally send wxCommandEvent-derived
events only for the user-generated events. The only @b exceptions to this rule are:

@li wxNotebook::AddPage: No event-free alternatives
@li wxNotebook::AdvanceSelection: No event-free alternatives
@li wxNotebook::DeletePage: No event-free alternatives
@li wxNotebook::SetSelection: Use wxNotebook::ChangeSelection instead, as
    wxNotebook::SetSelection is deprecated
@li wxTreeCtrl::Delete: No event-free alternatives
@li wxTreeCtrl::DeleteAllItems: No event-free alternatives
@li wxTreeCtrl::EditLabel: No event-free alternatives
@li All wxTextCtrl methods

wxTextCtrl::ChangeValue can be used instead of wxTextCtrl::SetValue but the other
functions, such as wxTextCtrl::Replace or wxTextCtrl::WriteText don't have event-free
equivalents.



@section overview_eventhandling_pluggable Pluggable Event Handlers

In fact, you don't have to derive a new class from a window class
if you don't want to. You can derive a new class from wxEvtHandler instead,
defining the appropriate event table, and then call wxWindow::SetEventHandler
(or, preferably, wxWindow::PushEventHandler) to make this
event handler the object that responds to events. This way, you can avoid
a lot of class derivation, and use instances of the same event handler class (but different
objects as the same event handler object shouldn't be used more than once) to
handle events from instances of different widget classes.

If you ever have to call a window's event handler
manually, use the GetEventHandler function to retrieve the window's event handler and use that
to call the member function. By default, GetEventHandler returns a pointer to the window itself
unless an application has redirected event handling using SetEventHandler or PushEventHandler.

One use of PushEventHandler is to temporarily or permanently change the
behaviour of the GUI. For example, you might want to invoke a dialog editor
in your application that changes aspects of dialog boxes. You can
grab all the input for an existing dialog box, and edit it 'in situ',
before restoring its behaviour to normal. So even if the application
has derived new classes to customize behaviour, your utility can indulge
in a spot of body-snatching. It could be a useful technique for on-line
tutorials, too, where you take a user through a serious of steps and
don't want them to diverge from the lesson. Here, you can examine the events
coming from buttons and windows, and if acceptable, pass them through to
the original event handler. Use PushEventHandler/PopEventHandler
to form a chain of event handlers, where each handler processes a different
range of events independently from the other handlers.



@section overview_eventhandling_winid Window Identifiers

Window identifiers are integers, and are used to
uniquely determine window identity in the event system (though you can use it
for other purposes). In fact, identifiers do not need to be unique
across your entire application just so long as they are unique within a
particular context you're interested in, such as a frame and its children. You
may use the @c wxID_OK identifier, for example, on any number of dialogs so
long as you don't have several within the same dialog.

If you pass @c wxID_ANY to a window constructor, an identifier will be
generated for you automatically by wxWidgets. This is useful when you don't
care about the exact identifier either because you're not going to process the
events from the control being created at all or because you process the events
from all controls in one place (in which case you should specify @c wxID_ANY
in the event table or wxEvtHandler::Connect call
as well. The automatically generated identifiers are always negative and so
will never conflict with the user-specified identifiers which must be always
positive.

See @ref page_stdevtid for the list of standard identifiers availabel.
You can use wxID_HIGHEST to determine the number above which it is safe to
define your own identifiers. Or, you can use identifiers below wxID_LOWEST.
Finally, you can allocate identifiers dynamically using wxNewId() function to.
If you use wxNewId() consistently in your application, you can be sure that
the your identifiers don't conflict accidentally.


@section overview_eventhandling_custom Custom Event Summary

@subsection overview_eventhandling_custom_general General approach

Since version 2.2.x of wxWidgets, each event type is identified by ID which
is given to the event type @e at runtime which makes it possible to add
new event types to the library or application without risking ID clashes
(two different event types mistakingly getting the same event ID). This
event type ID is stored in a struct of type @b const wxEventType.

In order to define a new event type, there are principally two choices.
One is to define a entirely new event class (typically deriving from
wxEvent or wxCommandEvent.

The other is to use the existing event classes and give them an new event
type. You'll have to define and declare a new event type using either way,
and this is done using the following macros:

@code
// in the header of the source file
BEGIN_DECLARE_EVENT_TYPES()
DECLARE_EVENT_TYPE(name, value)
END_DECLARE_EVENT_TYPES()

// in the implementation
DEFINE_EVENT_TYPE(name)
@endcode

You can ignore the @e value parameter of the DECLARE_EVENT_TYPE macro
since it is used only for backwards compatibility with wxWidgets 2.0.x based
applications where you have to give the event type ID an explicit value.
See also the @ref page_samples_event for an example of code
defining and working with the custom event types.


@subsection overview_eventhandling_custom_existing Using Existing Event Classes

If you just want to use a wxCommandEvent with
a new event type, you can then use one of the generic event table macros
listed below, without having to define a new macro yourself. This also
has the advantage that you won't have to define a new wxEvent::Clone()
method for posting events between threads etc. This could look like this
in your code:

@code
DECLARE_EVENT_TYPE(wxEVT_MY_EVENT, -1)
DEFINE_EVENT_TYPE(wxEVT_MY_EVENT)

// user code intercepting the event

BEGIN_EVENT_TABLE(MyFrame, wxFrame)
EVT_MENU    (wxID_EXIT, MyFrame::OnExit)
// ....
EVT_COMMAND  (ID_MY_WINDOW, wxEVT_MY_EVENT, MyFrame::OnMyEvent)
END_EVENT_TABLE()

void MyFrame::OnMyEvent( wxCommandEvent  )
{
    // do something
    wxString text = event.GetText();
}


// user code sending the event

void MyWindow::SendEvent()
{
    wxCommandEvent event( wxEVT_MY_EVENT, GetId() );
    event.SetEventObject( this );
    // Give it some contents
    event.SetText( wxT("Hallo") );
    // Send it
    GetEventHandler()->ProcessEvent( event );
}
@endcode


@subsection overview_eventhandling_custom_generic Generic Event Table Macros

@beginTable
@row2col{EVT_CUSTOM(event\, id\, func),
        Allows you to add a custom event table
        entry by specifying the event identifier (such as wxEVT_SIZE),
        the window identifier, and a member function to call.}
@row2col{EVT_CUSTOM_RANGE(event\, id1\, id2\, func),
        The same as EVT_CUSTOM, but responds to a range of window identifiers.}
@row2col{EVT_COMMAND(id\, event\, func),
        The same as EVT_CUSTOM, but expects a member function with a
        wxCommandEvent argument.}
@row2col{EVT_COMMAND_RANGE(id1\, id2\, event\, func),
        The same as EVT_CUSTOM_RANGE, but
        expects a member function with a wxCommandEvent argument.}
@row2col{EVT_NOTIFY(event\, id\, func),
        The same as EVT_CUSTOM, but
        expects a member function with a wxNotifyEvent argument.}
@row2col{EVT_NOTIFY_RANGE(event\, id1\, id2\, func),
        The same as EVT_CUSTOM_RANGE, but
        expects a member function with a wxNotifyEvent argument.}
@endTable


@subsection overview_eventhandling_custom_ownclass Defining Your Own Event Class

Under certain circumstances, it will be required to define your own event
class e.g. for sending more complex data from one place to another. Apart
from defining your event class, you will also need to define your own
event table macro (which is quite long). Watch out to put in enough
casts to the inherited event function. Here is an example:

@code
// code defining event

class wxPlotEvent: public wxNotifyEvent
{
public:
    wxPlotEvent( wxEventType commandType = wxEVT_NULL, int id = 0 );

    // accessors
    wxPlotCurve *GetCurve()
        { return m_curve; }

    // required for sending with wxPostEvent()
    virtual wxEvent *Clone() const;

private:
    wxPlotCurve   *m_curve;
};

DECLARE_EVENT_TYPE( wxEVT_PLOT_ACTION, -1 )

typedef void (wxEvtHandler::*wxPlotEventFunction)(wxPlotEvent&);

#define EVT_PLOT(id, fn) \
    DECLARE_EVENT_TABLE_ENTRY( wxEVT_PLOT_ACTION, id, -1, \
    (wxObjectEventFunction) (wxEventFunction) (wxCommandEventFunction) (wxNotifyEventFunction) \
    wxStaticCastEvent( wxPlotEventFunction, &fn ), (wxObject *) NULL ),


// code implementing the event type and the event class

DEFINE_EVENT_TYPE( wxEVT_PLOT_ACTION )

wxPlotEvent::wxPlotEvent( ...


// user code intercepting the event

BEGIN_EVENT_TABLE(MyFrame, wxFrame)
EVT_PLOT  (ID_MY_WINDOW, MyFrame::OnPlot)
END_EVENT_TABLE()

void MyFrame::OnPlot( wxPlotEvent &event )
{
    wxPlotCurve *curve = event.GetCurve();
}


// user code sending the event

void MyWindow::SendEvent()
{
    wxPlotEvent event( wxEVT_PLOT_ACTION, GetId() );
    event.SetEventObject( this );
    event.SetCurve( m_curve );
    GetEventHandler()->ProcessEvent( event );
}
@endcode


@section overview_eventhandling_macros Event Handling Summary

For the full list of event classes, please see the
@ref group_class_events "event classes group page".


@todo for all controls state clearly when calling a member function results in an 
      event being generated and when it doesn't (possibly updating also the 
      'Events generated by the user vs programmatically generated events' paragraph 
      of the 'Event handling overview' with the list of the functions which break 
      that rule).

*/