fltk/documentation/subclassing.html
Bill Spitzak 4c53a5d8f4 Added optimization for SGI builds (mike: please run autoconf before making
a distribution).
Documentation fixes.


git-svn-id: file:///fltk/svn/fltk/trunk@259 ea41ed52-d2ee-0310-a9c1-e6b18d33e121
1999-01-31 07:43:16 +00:00

396 lines
17 KiB
HTML

<HTML><BODY>
<H1 ALIGN=RIGHT><A NAME=subclassing>7 - Adding and Extending Widgets</A></H1>
This chapter describes how to add your own widgets or extend existing
widgets in FLTK.
<H2>Subclassing</H2>
New widgets are created by <I>subclassing</I> an existing FLTK widget,
typically <TT>Fl_Widget</TT> for controls and <TT>Fl_Group</TT> for
composite widgets.
<P>A control widget typically interacts with the user to receive and/or
display a value of some sort. </P>
<P>A composite widget widget holds a list of child widgets and handles moving,
sizing, showing, or hiding them as needed. <TT>Fl_Group</TT> is the
main composite widget widget class in FLTK, and all of the other composite widgets (<TT>
Fl_Pack</TT>, <TT>Fl_Scroll</TT>, <TT>Fl_Tabs</TT>, <TT>Fl_Tile</TT>,
and <TT>Fl_Window</TT>) are subclasses of it. </P>
<P>You can also subclass other existing widgets to provide a different
look or user-interface. For example, the button widgets are all
subclasses of <TT>Fl_Button</TT> since they all interact with the user
via a mouse button click. The only difference is the code that draws
the face of the button. </P>
<H2>Making a Subclass of Fl_Widget</H2>
Your subclasses can directly descend from <TT>Fl_Widget</TT> or any
subclass of <TT>Fl_Widget</TT>. <TT>Fl_Widget</TT> has only four
virtual methods, and overriding some or all of these may be necessary.
<H2>The Constructor</H2>
The constructor should have the following arguments:
<UL>
<PRE>
MyClass(int x, int y, int w, int h, const char *label = 0);
</PRE>
</UL>
This will allow the class to be used in <A href=fluid.html#fluid>FLUID</A>
without problems.
<P>The constructor must call the constructor for the base class and
pass the same arguments: </P>
<UL>
<PRE>
MyClass::MyClass(int x, int y, int w, int h, const char *label)
: Fl_Widget(x, y, w, h, label) {
// do initialization stuff...
}
</PRE>
</UL>
<TT>Fl_Widget</TT>'s protected constructor sets <TT>x()</TT>, <TT>y()</TT>,
<TT>w()</TT>, <TT>h()</TT>, and <TT>label()</TT> to the passed values
and initializes the other instance variables to:
<UL>
<PRE>
type(0);
box(FL_NO_BOX);
color(FL_GRAY);
selection_color(FL_GRAY);
labeltype(FL_NORMAL_LABEL);
labelstyle(FL_NORMAL_STYLE);
labelsize(FL_NORMAL_SIZE);
labelcolor(FL_BLACK);
align(FL_ALIGN_CENTER);
callback(default_callback,0);
flags(ACTIVE|VISIBLE);
</PRE>
</UL>
<H2>Protected Methods of Fl_Widget</H2>
The following methods are provided for subclasses to use:
<UL>
<LI><A href=#clear_visible><TT>Fl_Widget::clear_visible</TT></A></LI>
<LI><A href=#damage><TT>Fl_Widget::damage</TT></A></LI>
<LI><A href=#draw_box><TT>Fl_Widget::draw_box</TT></A></LI>
<LI><A href=#draw_label><TT>Fl_Widget::draw_label</TT></A></LI>
<LI><A href=#set_flag><TT>Fl_Widget::set_flag</TT></A></LI>
<LI><A href=#set_visible><TT>Fl_Widget::set_visible</TT></A></LI>
<LI><A href=#test_shortcut><TT>Fl_Widget::test_shortcut</TT></A></LI>
<LI><A href=#type><TT>Fl_Widget::type</TT></A></LI>
</UL>
<H4><A name=damage>void Fl_Widget::damage(uchar mask)
<BR> void Fl_Widget::damage(uchar mask, int x, int y, int w, int h)
<BR> uchar Fl_Widget::damage()</A></H4>
The first form indicates that a partial update of the object is
needed. The bits in mask are OR'd into <TT>damage()</TT>. Your <TT>
draw()</TT> routine can examine these bits to limit what it is
drawing. The public method <TT>Fl_Widget::redraw()</TT> simply does
<TT> Fl_Widget::damage(FL_DAMAGE_ALL)</TT>, but the implementation of
your widget can call the private <TT>damage(n)</TT>.
<P>The second form indicates that a region is damaged. If only these
calls are done in a window (no calls to <TT>damage(n)</TT>) then FLTK
will clip to the union of all these calls before drawing anything.
This can greatly speed up incremental displays. The mask bits are
OR'd into <TT>damage()</TT> unless this is a <TT>Fl_Window</TT> widget. </P>
<P>The third form returns the bitwise-OR of all <TT>damage(n)</TT>
calls done since the last <TT>draw()</TT>.</P>
<P><I>When redrawing your widgets you should look at the damage bits to
see what parts of your widget need redrawing.</I> The <tt>handle()</tt>
method can then set individual damage bits to limit the amount of drawing
that needs to be done:
<UL><PRE>
MyClass::handle(int event) {
...
if (change_to_part1) damage(1);
if (change_to_part2) damage(2);
if (change_to_part3) damage(4);
}
MyClass::draw() {
if (damage() & FL_DAMAGE_ALL) {
... draw frame/box and other static stuff ...
}
if (damage() & (FL_DAMAGE_ALL | 1)) draw_part1();
if (damage() & (FL_DAMAGE_ALL | 2)) draw_part2();
if (damage() & (FL_DAMAGE_ALL | 4)) draw_part3();
}
</PRE></UL>
<H4><A name=draw_box>void Fl_Widget::draw_box() const
<BR></A>void Fl_Widget::draw_box(Fl_Boxtype b, ulong c) const</H4>
The first form draws this widget's <TT>box()</TT>, using the
dimensions of the widget. The second form uses <TT>b</TT> as the box
type and <TT>c</TT> as the color for the box.
<H4><A name=draw_label>void Fl_Widget::draw_label() const
<BR> void Fl_Widget::draw_label(int x, int y, int w, int h) const
<BR> void Fl_Widget::draw_label(int x, int y, int w, int h, Fl_Align
align) const</A></H4>
This is the usual function for a <TT>draw()</TT> method to call to
draw the widget's label. It does not draw the label if it is supposed
to be outside the box (on the assumption that the enclosing group will
draw those labels).
<P>The second form uses the passed bounding box instead of the widget's
bounding box. This is useful so &quot;centered&quot; labels are aligned with some
feature, like a moving slider. </P>
<P>The third form draws the label anywhere. It acts as though <TT>
FL_ALIGN_INSIDE</TT> has been forced on so the label will appear inside
the passed bounding box. This is designed for parent groups to draw
labels with. </P>
<H4><A name=set_flag>void Fl_Widget::set_flag(SHORTCUT_LABEL)</A></H4>
Modifies <TT>draw_label()</TT> so that '&amp;' characters cause an underscore
to be printed under the next letter.
<H4><A name=set_visible>void Fl_Widget::set_visible()</A>
<BR><A name=clear_visible>void Fl_Widget::clear_visible()</A></H4>
Fast inline versions of <TT>Fl_Widget::hide()</TT> and <TT>
Fl_Widget::show()</TT>. These do not send the <TT>FL_HIDE</TT> and <TT>
FL_SHOW</TT> events to the widget.
<H4><A name=test_shortcut>int Fl_Widget::test_shortcut() const
<BR> static int Fl_Widget::test_shortcut(const char *s)</A></H4>
The first version tests <TT>Fl_Widget::label()</TT> against the
current event (which should be a <TT>FL_SHORTCUT</TT> event). If the
label contains a '&amp;' character and the character after it matches the key
press, this returns true. This returns false if the <TT>SHORTCUT_LABEL</TT>
flag is off, if the label is <TT>NULL</TT> or does not have a
'&amp;' character in it, or if the keypress does not match the character.
<P>The second version lets you do this test against an arbitrary
string. </P>
<H4><A name=type>uchar Fl_Widget::type() const
<BR> void Fl_Widget::type(uchar t)</A></H4>
The property <TT>Fl_Widget::type()</TT> can return an arbitrary 8-bit
identifier, and can be set with the protected method <TT>type(uchar t)</TT>
. This value had to be provided for Forms compatibility, but you can
use it for any purpose you want. Try to keep the value less than 100
to not interfere with reserved values.
<P>FLTK does not use RTTI (Run Time Typing Infomation), to enhance
portability. But this may change in the near future if RTTI becomes
standard everywhere. </P>
<P>If you don't have RTTI you can use the clumsy FLTK mechanisim, by
having <TT>type()</TT> use a unique value. These unique values must
be greater than the symbol <TT>FL_RESERVED_TYPE</TT> (which is 100).
Look through the header files for <TT>FL_RESERVED_TYPE</TT> to find an
unused number. If you make a subclass of <TT>Fl_Window</TT>
you must use <TT>FL_WINDOW + n</TT> (<TT>n</tt> must be in the
range 1 to 7). </P>
<H2>Handling Events</H2>
The virtual method <TT>int Fl_Widget::handle(int event)</TT> is called
to handle each event passed to the widget. It can:
<UL>
<LI>Change the state of the widget. </LI>
<LI>Call <A href=Fl_Widget.html#Fl_Widget.redraw><TT>Fl_Widget::redraw()</TT>
</A> if the widget needs to be redisplayed. </LI>
<LI>Call <A href=Fl_Widget.html#Fl_Widget.damage><TT>
Fl_Widget::damage(n)</TT></A> if the widget needs a partial-update
(assumming you provide support for this in your <TT>Fl_Widget::draw()</TT>
method). </LI>
<LI>Call <A href=Fl_Widget.html#Fl_Widget.do_callback><TT>
Fl_Widget::do_callback()</TT></A> if a callback should be generated. </LI>
<LI>Call <TT>Fl_Widget::handle()</TT> on child widgets. </LI>
</UL>
Events are identified by the integer argument. Other information
about the most recent event is stored in static locations and aquired
by calling the <A href=events.html#events><TT>Fl::event_*()</TT></A>
functions. This information remains valid until another event is
handled.
<P>Here is a sample <TT>handle()</TT> method for a widget that acts as
a pushbutton and also accepts the keystroke 'x' to cause the callback: </P>
<UL>
<PRE>
int MyClass::handle(int event) {
switch(event) {
case FL_PUSH:
highlight = 1;
redraw();
return 1;
case FL_DRAG: {
int t = Fl::event_inside(this);
if (t != highlight) {
highlight = t;
redraw();
}
}
return 1;
case FL_RELEASE:
if (highlight) {
highlight = 0;
redraw();
do_callback();
// never do anything after a callback, as the callback
// may delete the widget!
}
return 1;
case FL_SHORTCUT:
if (Fl::event_key() == 'x') {
do_callback();
return 1;
}
return 0;
default:
return 0;
}
}
</PRE>
</UL>
You must return non-zero if your <TT>handle()</TT> method uses the
event. If you return zero it indicates to the parent widget that it can
try sending the event to another widget.
<H2>Drawing the Widget</H2>
The <TT>draw()</TT> virtual method is called when FLTK wants you to
redraw your widget. It will be called if and only if <TT>damage()</TT>
is non-zero, and <TT>damage()</TT> will be cleared to zero after it
returns. <TT>draw()</TT> should be declared protected, so that it can't
be called from non-drawing code.
<P><TT>damage()</TT> contains the bitwise-OR of all the <TT>damage(n)</TT>
calls to this widget since it was last drawn. This can be used for
minimal update, by only redrawing the parts whose bits are set. FLTK
will turn on the <TT>FL_DAMAGE_ALL</TT> bit if it thinks the entire widget
must be redrawn (e.g. for an expose event). </P>
<P>Expose events (and the above <TT>damage(b,x,y,w,h)</TT>) will cause <TT>
draw()</TT> to be called with FLTK's <A href=drawing.html#clipping>
clipping</A> turned on. You can greatly speed up redrawing in some
cases by testing <TT>fl_not_clipped(x,y,w,h)</TT> or <TT>fl_clip_box(...)</TT> and
skipping invisible parts. </P>
<P>Besides the protected methods described above, FLTK provides a large
number of basic drawing functions, which are described <A href=drawing.html#drawing>
below</A>. </P>
<H2>Resizing the Widget</H2>
The <TT>resize(int x, int y, int w, int h)</TT> method is called when
the widget is being resized or moved. The arguments are the new
position, width, and height. <TT>x()</TT>, <TT>y()</TT>, <TT>w()</TT>,
and <TT>h()</TT> still remain the old size. You must call <TT>resize()</TT>
on your base class with the same arguments to get the widget size to
actually change.
<P>This should <I>not</I> call <TT>redraw()</TT>, at least if only the <TT>
x()</TT> and <TT>y()</TT> change. This is because composite widgets like <A href=Fl_Scroll.html#Fl_Scroll>
<TT>Fl_Scroll</TT></A> may have a more efficient way of drawing the new
position. </P>
<H2>Making a Composite Widget</H2>
A &quot;composite&quot; widget contains one or more &quot;child&quot; widgets.
To make a composite widget you should subclass <A href=Fl_Group.html#Fl_Group><TT>Fl_Group</TT></A>
. It is possible to make a composite object that is not a subclass of <TT>
Fl_Group</TT>, but you'll have to duplicate the code in <TT>Fl_Group</TT>
anyways.
<P>Instances of the child widgets may be included in the parent: </P>
<UL>
<PRE>
class MyClass : public Fl_Group {
Fl_Button the_button;
Fl_Slider the_slider;
...
};
</PRE>
</UL>
The constructor has to initialize these instances. They are
automatically <TT>add()</TT>ed to the group, since the <TT>Fl_Group</TT>
constructor does <TT>begin()</TT>. <I>Don't forget to call <TT>end()</TT>
or use the <A href=Fl_End.html#Fl_End><TT>Fl_End</TT></A> pseudo-class:</I>
<UL>
<PRE>
MyClass::MyClass(int x, int y, int w, int h) :
Fl_Group(x, y, w, h),
the_button(x + 5, y + 5, 100, 20),
the_slider(x, y + 50, w, 20)
{
...(you could add dynamically created child widgets here)...
end(); // don't forget to do this!
}
</PRE>
</UL>
The child widgets need callbacks. These will be called with a pointer
to the children, but the widget itself may be found in the <TT>parent()</TT>
pointer of the child. Usually these callbacks can be static private
methods, with a matching private method:
<UL>
<PRE>
void MyClass::slider_cb(Fl_Widget* v, void *) { // static method
((MyClass*)(v-&gt;parent())-&gt;slider_cb();
}
void MyClass::slider_cb() { // normal method
use(the_slider-&gt;value());
}
</PRE>
</UL>
If you make the <TT>handle()</TT> method, you can quickly pass all the
events to the children using the <TT>Fl_Group::handle()</TT> method.
You don't need to override <TT>handle()</TT> if your composite widget
does nothing other than pass events to the children:
<UL>
<PRE>
int MyClass::handle(int event) {
if (Fl_Group::handle(event)) return 1;
... handle events that children don't want ...
}
</PRE>
</UL>
If you override <TT>draw()</TT> you need to draw all the children. If <TT>
redraw()</TT> or <TT>damage()</TT> is called on a child, <TT>
damage(FL_DAMAGE_CHILD)</TT> is done to the group, so this bit of <TT>
damage()</TT> can be used to indicate that a child needs to be drawn.
It is fastest if you avoid drawing anything else in this case:
<UL>
<PRE>
int MyClass::draw() {
Fl_Widget *const*a = array();
if (damage() == FL_DAMAGE_CHILD) { // only redraw some children
for (int i = children(); i --; a ++) update_child(**a);
} else { // total redraw
... draw background graphics ...
// now draw all the children atop the background:
for (int i = children_; i --; a ++) {
draw_child(**a);
draw_outside_label(**a); // you may not want to do this
}
}
}
</PRE>
</UL>
<TT>Fl_Group</TT> provides some protected methods to make drawing
easier:
<UL>
<LI><A href=#draw_child>draw_child</A></LI>
<LI><A href=#draw_outside_label>draw_outside_label</A></LI>
<LI><A href=#update_child>update_child</A></LI>
</UL>
<H4><A name=draw_child>void Fl_Group::draw_child(Fl_Widget&amp;)</A></H4>
This will force the child's <TT>damage()</TT> bits all to one and call <TT>
draw()</TT> on it, then clear the <TT>damage()</TT>. You should call
this on all children if a total redraw of your widget is requested, or
if you draw something (like a background box) that damages the child.
Nothing is done if the child is not <TT>visible()</TT> or if it is
clipped.
<H4><A name=draw_outside_label>void
Fl_Group::draw_outside_label(Fl_Widget&amp;) const</A></H4>
Draw the labels that are <I>not</I> drawn by <A href=#draw_label><TT>
draw_label()</TT></A>. If you want more control over the label
positions you might want to call <TT>child-&gt;draw_label(x,y,w,h,a)</TT>.
<H4><A name=update_child>void Fl_Group::update_child(Fl_Widget&amp;)</A></H4>
Draws the child only if its <TT>damage()</TT> is non-zero. You
should call this on all the children if your own damage is equal to
FL_DAMAGE_CHILD. Nothing is done if the child is not <TT>visible()</TT>
or if it is clipped.
<H2>Cut and Paste Support</H2>
FLTK provides routines to cut and paste ASCII text (in the future this
may be UTF-8) between applications:
<UL>
<LI><A href=functions.html#paste><TT>Fl::paste</TT></A></LI>
<LI><A href=functions.html#selection><TT>Fl::selection</TT></A></LI>
<LI><A href=#selection_length><TT>Fl::selection_length</TT></A></LI>
<LI><A href=functions.html#selection_owner><TT>Fl::selection_owner</TT></A></LI>
</UL>
It may be possible to cut/paste non-ASCII data by using <A href=osissues.html#add_handler>
<TT>Fl::add_handler()</TT></A>.
<H2>Making a subclass of Fl_Window</H2>
You may want your widget to be a subclass of <TT>Fl_Window</TT>. This
can be useful if your widget wants to occupy an entire window, and can
also be used to take advantage of system-provided clipping, or to work
with a library that expects a system window ID to indicate where to
draw.
<P>Subclassing <TT>Fl_Window </TT>is almost exactly like subclassing <TT>
Fl_Widget</TT>, and in fact you can easily switch a subclass back and
forth. Watch out for the following differences: </P>
<OL>
<LI><TT>Fl_Window</TT> is a subclass of <TT>Fl_Group</TT> so <I>make
sure your constructor calls <TT>end()</TT></I> (unless you actually
want children added to your window). </LI>
<LI>When handling events and drawing, the upper-left corner is at 0,0,
not <TT>x(),y()</TT> as in other <TT>Fl_Widget</TT>'s. For instance, to
draw a box around the widget, call <TT>draw_box(0, 0, w(), h())</TT>,
rather than <TT>draw_box(x(), y(), w(), h())</TT>. </LI>
</OL>
You may also want to subclass <TT>Fl_Window</TT> in order to get
access to different visuals or to change other attributes of the
windows. See <A href=osissues.html#osissues>"Appendix F - Operating
System Issues"</A> for more information. </BODY></HTML>