This chapter shows how to use the Fast Light User-Interface Designer ("FLUID") to create your GUIs.
The Fast Light User Interface Designer, or FLUID, is a graphical editor that is used to produce FLTK source code. FLUID edits and saves its state in
.fl files. These files are text, and you can (with care) edit them in a text editor, perhaps to get some special effects.
FLUID can "compile" the
.fl file into a
.cxx and a
.h file. The
.cxx file defines all the objects from the
.fl file and the
.h file declares all the global ones. FLUID also supports localization (Internationalization) of label strings using message files and the GNU gettext or POSIX catgets interfaces.
A simple program can be made by putting all your code (including a
main() function) into the
.fl file and thus making the
.cxx file a single source file to compile. Most programs are more complex than this, so you write other
.cxx files that call the FLUID functions. These
.cxx files must
.h file or they can
.cxx file so it still appears to be a single source file.
Normally the FLUID file defines one or more functions or classes which output C++ code. Each function defines one or more FLTK windows, and all the widgets that go inside those windows.
Widgets created by FLUID are either "named", "complex named" or "unnamed". A named widget has a legal C++ variable identifier as its name (i.e. only alphanumeric and underscore). In this case FLUID defines a global variable or class member that will point at the widget after the function defining it is called. A complex named object has punctuation such as
'->' or any other symbols in its name. In this case FLUID assigns a pointer to the widget to the name, but does not attempt to declare it. This can be used to get the widgets into structures. An unnamed widget has a blank name and no pointer is stored.
Widgets may either call a named callback function that you write in another source file, or you can supply a small piece of C++ source and FLUID will write a private callback function into the
To run FLUID under UNIX, type:
to edit the
filename.fl. If the file does not exist you will get an error pop-up, but if you dismiss it you will be editing a blank file of that name. You can run FLUID without any name, in which case you will be editing an unnamed blank setup (but you can use save-as to write it to a file).
You can provide any of the standard FLTK switches before the filename:
Changing the colors may be useful to see what your interface will look at if the user calls it with the same switches. Similarly, using "-scheme plastic" will show how the interface will look using the "plastic" scheme.
In the current version, if you don't put FLUID into the background with
'&' then you will be able to abort FLUID by typing
CTRL-C on the terminal. It will exit immediately, losing any changes.
To run FLUID under WIN32, double-click on the FLUID.exe file. You can also run FLUID from the Command Prompt window. FLUID always runs in the background under WIN32.
FLUID can also be called as a command-line "compiler" to create the
.h file from a
.fl file. To do this type:
This is the same as the menu 'File/Write Code...'. It will read the
filename.fl file and write
filename.h. Any leading directory on
filename.fl will be stripped, so they are always written to the current directory. If there are any errors reading or writing the files, FLUID will print the error and exit with a non-zero code. You can use the following lines in a makefile to automate the creation of the source and header files:
Most versions of make support rules that cause
.fl files to be compiled:
If you use
FLUID will also write the "strings" for internationalization in file 'filename.txt' (menu: 'File/Write Strings...').
Finally there is another option which is useful for program developers who have many .fl files and want to upgrade them to the current FLUID version. FLUID will read the
filename.fl file, save it, and exit immediately. This writes the file with current syntax and options and the current FLTK version in the header of the file. Use
filename.fl . You may combine this with '-c' or '-cs'.
FLUID is an amazingly powerful little program. However, this power comes at a price as it is not always obvious how to accomplish seemingly simple tasks with it. This tutorial will show you how to generate a complete user interface class with FLUID that is used for the CubeView program provided with FLTK.
The window is of class CubeViewUI, and is completely generated by FLUID, including class member functions. The central display of the cube is a separate subclass of Fl_Gl_Window called CubeView. CubeViewUI manages CubeView using callbacks from the various sliders and rollers to manipulate the viewing angle and zoom of CubeView.
At the completion of this tutorial you will (hopefully) understand how to:
The CubeView class is a subclass of Fl_Gl_Window. It has methods for setting the zoom, the x and y pan, and the rotation angle about the x and y axes.
You can safely skip this section as long as you realize that CubeView is a sublass of Fl_Gl_Window and will respond to calls from CubeViewUI, generated by FLUID.
Here is the CubeView class definition, as given by its header file "test/CubeView.h":
Here is the CubeView implementation. It is very similar to the "cube" demo included with FLTK.
We will completely construct a window to display and control the CubeView defined in the previous section using FLUID.
Once you have started FLUID, the first step in defining a class is to create a new class within FLUID using the New->Code->Class menu item. Name the class "CubeViewUI" and leave the subclass blank. We do not need any inheritance for this window. You should see the new class declaration in the FLUID browser window.
Click on the CubeViewUI class in the FLUID window and add a new method by selecting New->Code->Function/Method. The name of the function will also be CubeViewUI. FLUID will understand that this will be the constructor for the class and will generate the appropriate code. Make sure you declare the constructor public.
Then add a window to the CubeViewUI class. Highlight the name of the constructor in the FLUID browser window and click on New->Group->Window. In a similar manner add the following to the CubeViewUI constructor:
None of these additions need be public. And they shouldn't be unless you plan to expose them as part of the interface for CubeViewUI.
When you are finished you should have something like this:
We will talk about the
show() method that is highlighted shortly.
What we have is nice, but does little to show our cube. We have already defined the CubeView class and we would like to show it within the CubeViewUI.
The CubeView class inherits the Fl_Gl_Window class, which is created in the same way as a Fl_Box widget. Use New->Other->Box to add a square box to the main window. This will be no ordinary box, however.
The Box properties window will appear. The key to letting CubeViewUI display CubeView is to enter CubeView in the Class: text entry box. This tells FLUID that it is not an Fl_Box, but a similar widget with the same constructor.
In the Extra Code: field enter
#include is important, as we have just included CubeView as a member of CubeViewUI, so any public CubeView methods are now available to CubeViewUI.
Each of the widgets we defined before adding CubeView can have callbacks that call CubeView methods. You can call an external function or put in a short amount of code in the Callback field of the widget panel. For example, the callback for the
ypan slider is:
cube->redraw() after changing the value to update the CubeView window. CubeView could easily be modified to do this, but it is nice to keep this exposed. In the case where you may want to do more than one view change only redrawing once saves a lot of time.
There is no reason to wait until after you have added CubeView to enter these callbacks. FLUID assumes you are smart enough not to refer to members or functions that don't exist.
You can add class methods within FLUID that have nothing to do with the GUI. As an example add a show function so that CubeViewUI can actually appear on the screen.
Make sure the top level CubeViewUI is selected and select New->Code->Function/Method. Just use the name
show(). We don't need a return value here, and since we will not be adding any widgets to this method FLUID will assign it a return type of
Once the new method has been added, highlight its name and select New->Code->Code. Enter the method's code in the code window.
If you need to add code to initialize a class, for example setting initial values of the horizontal and vertical angles in the CubeView, you can simply highlight the constructor and select New->Code->Code. Add any required code.
Now that we have completely defined the CubeViewUI, we have to generate the code. There is one last trick to ensure this all works. Open the preferences dialog from Edit->Preferences.
At the bottom of the preferences dialog box is the key: "Include Header from Code". Select that option and set your desired file extensions and you are in business. You can include the CubeViewUI.h (or whatever extension you prefer) as you would any other C++ class.
The following sections describe each of the windows in FLUID.
The main window shows a menu bar and a scrolling browser of all the defined widgets. The name of the
.fl file being edited is shown in the window title.
The widgets are stored in a hierarchy. You can open and close a level by clicking the "triangle" at the left of a widget. The leftmost widgets are the parents, and all the widgets listed below them are their children. Parents don't have to have any children.
The top level of the hierarchy is composed of functions and classes. Each of these will produce a single C++ public function or class in the output
.cxx file. Calling the function or instantiating the class will create all of the child widgets.
The second level of the hierarchy contains the windows. Each of these produces an instance of class Fl_Window.
Below that are either widgets (subclasses of Fl_Widget) or groups of widgets (including other groups). Plain groups are for layout, navigation, and resize purposes. Tab groups provide the well-known file-card tab interface.
Widgets are shown in the browser by either their name (such as "main_panel" in the example), or by their type and label (such as "Button "the green"").
You select widgets by clicking on their names, which highlights them (you can also select widgets from any displayed window). You can select many widgets by dragging the mouse across them, or by using Shift+Click to toggle them on and off. To select no widgets, click in the blank area under the last widget. Note that hidden children may be selected even when there is no visual indication of this.
You open widgets by double-clicking on them, or (to open several widgets you have picked) by typing the F1 key. A control panel will appear so you can change the widget(s).
The menu bar at the top is duplicated as a pop-up menu on any displayed window. The shortcuts for all the menu items work in any window. The menu items are:
.flfile. You are asked for confirmation if you have changed the current file.
.fdfiles produced by the Forms and XForms "fdesign" programs. It is best to File/Merge them instead of opening them. FLUID does not understand everything in a
.fdfile, and will print a warning message on the controlling terminal for all data it does not understand. You will probably need to edit the resulting setup to fix these errors. Be careful not to save the file without changing the name, as FLUID will write over the
.fdfile with its own format, which fdesign cannot read!
.flfile, without changing the name of the current
.flfile. All the functions (even if they have the same names as the current ones) are added, and you will have to use cut/paste to put the widgets where you want.
.flfile. If the file is unnamed then FLUID will ask for a filename.
.hfile. These are exactly the same as the files you get when you run FLUID with the
.flfile, with the leading directory and trailing ".fl" stripped, and ".h" or ".cxx" appended.
.flfile, with the leading directory and trailing ".fl" stripped, and ".txt", ".po", or ".msg" appended depending on the Internationalization Mode.
.cxxoutput be a self-contained program that can be compiled and executed. This is done by deleting the function name so
main(argc,argv)is used. The function will call
show()on all the windows it creates and then call
Fl::run(). This can also be used to test resize behavior or other parts of the user interface.
When you double-click on a widget or a set of widgets you will get the "widget attribute panel".
When you change attributes using this panel, the changes are reflected immediately in the window. It is useful to hit the "no overlay" button (or type Ctrl+Shift+O) to hide the red overlay so you can see the widgets more accurately, especially when setting the box type.
If you have several widgets selected, they may have different values for the fields. In this case the value for one of the widgets is shown. But if you change this value, all of the selected widgets are changed to the new value.
Hitting "OK" makes the changes permanent. Selecting a different widget also makes the changes permanent. FLUID checks for simple syntax errors such as mismatched parenthesis in any code before saving any text.
"Revert" or "Cancel" put everything back to when you last brought up the panel or hit OK. However in the current version of FLUID, changes to "visible" attributes (such as the color, label, box) are not undone by revert or cancel. Changes to code like the callbacks are undone, however.
#includeheader file is put in the
.hfile. You must provide a
#includeline as the first line of the "Extra Code" which declares your subclass.
privatein the class. Otherwise it controls whether the widget is declared
staticor global (
#or the word
externthen FLUID thinks this is an "include" line, and it is written to the
.hfile. If the same include line occurs several times then only one copy is written.
o. The window being constructed is pointed to by the local variable
w. You can also access any arguments passed to the function here, and any named widgets that are before this one.
.cxxoutput file. The function prototype is
void name(class *o, void *v)so that you can refer to the widget as
v. FLUID will check for matching parenthesis, braces, and quotes, but does not do much other error checking. Be careful here, as it may be hard to figure out what widget is producing an error in the compiler.
user_data()of the widget. If blank the default value of zero is used. This can be any piece of C code that can be cast to a
void*in the callback function prototypes is replaced with this. You may want to use
longfor old XForms code. Be warned that anything other than
void*is not guaranteed to work! However on most architectures other pointer types are ok, and
longis usually ok, too.
when()field that are not in the menu. You should use the extra code fields to put these values in.
Double-clicking a window name in the browser will display it, if not displayed yet. From this display you can select widgets, sets of widgets, and move or resize them. To close a window either double-click it or type
To select a widget, click it. To select several widgets drag a rectangle around them. Holding down shift will toggle the selection of the widgets instead.
You cannot pick hidden widgets. You also cannot choose some widgets if they are completely overlapped by later widgets. Use the browser to select these widgets.
The selected widgets are shown with a red "overlay" line around them. You can move the widgets by dragging this box. Or you can resize them by dragging the outer edges and corners. Hold down the Alt key while dragging the mouse to defeat the snap-to-grid effect for fine positioning.
If there is a tab box displayed you can change which child is visible by clicking on the file tabs. The child you pick is selected.
The arrow, tab, and shift+tab keys "navigate" the selection. Left, right, tab, or shift+tab move to the next or previous widgets in the hierarchy. Hit the right arrow enough and you will select every widget in the window. Up/down widgets move to the previous/next widgets that overlap horizontally. If the navigation does not seem to work you probably need to "Sort" the widgets. This is important if you have input fields, as FLTK uses the same rules when using arrow keys to move between input fields.
To "open" a widget, double click it. To open several widgets select them and then type F1 or pick "Edit/Open" off the pop-up menu.
Type Ctrl+o to temporarily toggle the overlay off without changing the selection, so you can see the widget borders.
You can resize the window by using the window manager border controls. FLTK will attempt to round the window size to the nearest multiple of the grid size and makes it big enough to contain all the widgets (it does this using illegal X methods, so it is possible it will barf with some window managers!). Notice that the actual window in your program may not be resizable, and if it is, the effect on child widgets may be different.
The panel for the window (which you get by double-clicking it) is almost identical to the panel for any other Fl_Widget. There are three extra items:
The contents of the image files in the Image and Inactive text fields are written to the
.cxx file. If many widgets share the same image then only one copy is written. Since the image data is embedded in the generated source code, you need only distribute the C++ code and not the image files themselves.
However, the filenames are stored in the
.fl file so you will need the image files as well to read the
.fl file. Filenames are relative to the location of the
.fl file and not necessarily the current directory. We recommend you either put the images in the same directory as the
.fl file, or use absolute path names.
libxpmlibrary. These files use C source code to define a pixmap. The filenames usually have the ".xpm" extension.
.giffile. Only the first image of an animated GIF file is used.
FLUID supports internationalization (I18N for short) of label strings used by widgets. The preferences window (
Ctrl+p) provides access to the I18N options.
FLUID supports three methods of I18N: use none, use GNU gettext, and use POSIX catgets. The "use none" method is the default and just passes the label strings as-is to the widget constructors.
The "GNU gettext" method uses GNU gettext (or a similar text-based I18N library) to retrieve a localized string before calling the widget constructor.
The "POSIX catgets" method uses the POSIX catgets function to retrieve a numbered message from a message catalog before calling the widget constructor.
FLUID's code support for GNU gettext is limited to calling a function or macro to retrieve the localized label; you still need to call
bindtextdomain() to select the appropriate language and message file.
To use GNU gettext for I18N, open the preferences window and choose "GNU gettext" from the Use: chooser. Two new input fields will then appear to control the include file and function/macro name to use when retrieving the localized label strings.
The #include field controls the header file to include for I18N; by default this is <libintl.h>, the standard I18N file for GNU gettext.
The Function: field controls the function (or macro) that will retrieve the localized message; by default the
gettext function will be called.
FLUID's code support for POSIX catgets allows you to use a global message file for all interfaces or a file specific to each
.fl file; you still need to call
setlocale() to select the appropriate language.
To use POSIX catgets for I18N, open the preferences window and choose "POSIX catgets" from the Use: chooser. Three new input fields will then appear to control the include file, catalog file, and set number for retrieving the localized label strings.
The #include field controls the header file to include for I18N; by default this is <nl_types.h>, the standard I18N file for POSIX catgets.
The File: field controls the name of the catalog file variable to use when retrieving localized messages; by default the file field is empty which forces a local (static) catalog file to be used for all of the windows defined in your
The Set: field controls the set number in the catalog file. The default set is 1 and rarely needs to be changed.
Declaration Blocks can be used to temporarily block out already designed code using
#if 0 and
#endif type construction. This will effectively avoid compilation of blocks of code. However, static code and data generated by this segment (menu items, images, include statements, etc.) will still be generated and likely cause compile-time warnings.
|[Prev] Using OpenGL||[Index]||Advanced FLTK [Next]|