D - Drawing Things in FLTK

This chapter covers the drawing functions that are provided with FLTK.

When Can You Draw Things in FLTK?

There are only certain places you can execute drawing code in FLTK. Calling these functions at other places will result in undefined behavior!

FLTK Drawing Functions

To use the drawing functions you must first include the <FL/fl_draw.H> header file. FLTK provides the following types of drawing functions:

Clipping

You can limit all your drawing to a region by calling fl_push_clip, and put the drawings back by using fl_pop_clip. Fltk may also set up clipping before draw() is called to limit the drawing to the region of the window that is damaged.

When drawing you can also test the current clip region with fl_not_clipped() and fl_clip_box(). By using these to skip over complex drawings that are clipped you can greatly speed up your program's redisplay.

The clip region is measured in pixels (it is unaffected by the current transformation matrix).

void fl_push_clip(int x, int y, int w, int h)

Pushes the intersection of the current region and this rectangle onto the clip stack.

void fl_clip_out(int x, int y, int w, int h)

Remove the rectangle from the current clip region, thus making it a more complex shape. This does not push the stack, it just replaces the top of it. This does not work on X or Win32 unless fl_clip() has been called at least once.

void fl_push_no_clip()

Pushes an empty clip region on the stack so nothing will be clipped. This lets you draw outside the current clip region. You should not use this :-)

void fl_pop_clip()

Restore the previous clip region. You must call fl_pop_clip() exactly once for every time you call fl_clip(). If you return to FLTK with the clip stack not empty unpredictable results occur.

int fl_not_clipped(int x, int y, int w, int h)

Return non-zero if the intersection of the rectangle and the current clip region is non-zero. If this returns zero you don't have to draw anything in that rectangle. Under X this returns 1 if the interesection is equal to the rectangle, and 2 if the intersection is only part of the rectangle.

int fl_clip_box(int x, int y, int w, int h, int &X, int &Y, int &W, int &H)

Find the smallest rectangle that surrounds the intersection of the rectangle x,y,w,h with the current clip region. This "bounding box" is returned in X,Y,W,H. The return value is non-zero if the bounding box is different than the rectangle. If the intersection is empty then W and H are set to zero.

This can be used to limit complex pixel operations (like drawing images) to the smallest rectangle needed to update the visible area.

Colors

void fl_color(Fl_Color)

Set the color for all subsequent drawing operations. Fl_Color is a typedef for a 32-bit integer containing r,g,b bytes and an "index" byte. The index is used if r,g,b is zero. For instance 0xFF008000 is 255 red, zero green, and 128 blue.

(On non-TrueColor X displays fltk rounds the desired color to the nearest color in a small (200) set of colors and allocates that from X to avoid consuming the entire colormap. On Windows colormapped displays the system dithering is used for all colors, which looks lousy, but Windows does not have the defective X behavior and thus you usually are not forced to set the screen to 8-bit mode)

Fl_Color fl_color()

Returns the last fl_color() that was set. This can be used for state save/restore.

Line dashes and thickness

void fl_line_style(int style, int width=0, char* dashes=0)

Set how to draw lines (the "pen"). If you change this it is your responsibility to set it back to the default with fl_line_style(0).

style is a bitmask in which you 'or' the following values. If you don't specify a dash type you will get a solid line. If you don't specify a cap or join type you will get a system-defined default of whatever value is fastest.

width is the number of pixels thick to draw the lines. Zero results in the system-defined default, which on both X and Windows is somewhat different and nicer than 1.

dashes is a pointer to an array of dash lengths, measured in pixels. The first location is how long to draw a solid portion, the next is how long to draw the gap, then the solid, etc. It is terminated with a zero-length entry. A null pointer or a zero-length array results in a solid line. Odd array sizes are not supported and result in undefined behavior. The dashes array does not work on Windows 95/98, use the dash styles instead.

Current Transformation

Path construction may be done with an arbitrary 2-D linear transformation. Many drawing operations ignore the transformation except for the integer translation, which ones do this may change in the future and on different platforms.

void fl_push_matrix()
void fl_pop_matrix()

Save and restore the current transformation. The maximum depth of the stack is 10.

void fl_scale(float x, float y)
void fl_scale(float x)
void fl_translate(float x, float y)
void fl_rotate(float d)
void fl_mult_matrix(float a, float b, float c, float d, float x, float y)

Concatenate another transformation onto the current one. The rotation angle is in degrees (not radians) and is counter-clockwise.

double fl_transform_x(double x, double y)
double fl_transform_y(double x, double y)

Return the actual x (or y) coordinate after x,y is transformed by the current transform. This value may be passed to fl_transformed_vertex or to the functions that dont do transforms yet, like fl_draw (but be warned that these may change in future versions).

double fl_transform_dx(double x, double y)
double fl_transform_dy(double x, double y)

Similar, but the translation is ignored, so this transforms a distance vector, instead of an actual point.

Path construction and drawing

These functions let you draw arbitrary shapes with 2-D linear transformations. The functionality matches that found in Adobe® PostScriptTM. On both X and WIN32 the transformed vertices are rounded to integers before drawing the line segments: this severely limits the accuracy of these functions for complex graphics, so use OpenGL when greater accuracy and/or performance is required.

void fl_newpath()

Clear the current "path". This is normally done by fl_fill() or any other drawing command.

void fl_vertex(float x, float y)

Add a single vertex to the current path. (If you are familiar with PostScript, this does a "moveto" if the path is clear or fl_closepath was done last, otherwise it does a "lineto").

void fl_vertex(int x, int y)

Add a single vertex to the current path. The current transformation, other than integer translation, may be ignored on some platforms, in order to make this faster! Use the double version if you want the transform to work.

void fl_transformed_vertex(float x, float y)

Add a single vertex to the current path. The passed values are not transformed and may be based on values returned by fl_transform_x(), fl_transform_y(), or fl_transform_dx() or fl_transform_dy().

void fl_closepath();

Similar to drawing another vertex back at the starting point, but fltk knows the path is closed. The next fl_vertex will start a new disconnected part of the shape.

It is harmless to call fl_closepath() several times in a row, or to call it before the first point. Sections with less than 3 points in them will not draw anything when filled.

void fl_curve(float x, float y, float x1, float y1, float x2, float y2, float x3, float y3)

Add a series of points on a Bezier curve to the path. The curve ends (and two of the points) are at x,y and x3,y3.

void fl_arc(float x, float y, float w, float h, float start, float end)

Add a series of points to the current path on the arc of an ellipse. The ellipse in inscribed in the x,y,w,h rectangle, and the start and end angles are measured in degrees counter-clockwise from 3 o'clock, 45 points at the upper-right corner of the rectangle. If end is less than start then it draws the arc in a clockwise direction.

void fl_ellipse(float x, float y, float w, float h)

Does closepath() and then adds a series of points on the edge of an ellipse inscribed in the given rectangle, then another closepath().

This tries to take advantage of the primitive drawing provided by X and Win32, which means it only draws the right thing if the rotation is a multiple of 90 degrees, or if the shape is a circle. Currently there can only be one ellipse or circle in a path.

void fl_circle(float x, float y, float r)

fl_circle() draws a circle of radius r centered on the point x,y. The result is always a circle, irregardless of scale. This also tries to take advantage of the X/Win32 graphics primitives like fl_ellipse.

void fl_points()

Draw a point (one pixel) for every vertex in the path, then clear the path.

void fl_stroke()

Draw a line between all the points in the path (see fl_line_type() for ways to set the thicknesss and dot pattern of the line).

void fl_fill()

Does fl_closepath() and then fill with the current color.

For portability, you should only draw polygons that appear the same whether "even/odd" or "non-zero" winding rules are used to fill them. This mostly means that holes should be drawn in the opposite direction of the outside.

void fl_fill_stroke(Fl_Color linecolor)

Does fl_fill(), then sets the current color to linecolor and does fl_stroke with the same closed path.

Drawing that bypasses the path mechanism

For speed and convienence, and to get some necessary graphics on the rather primitive interface provided by X and Win32, fltk has some calls that could be done with paths but are provided directly.

void fl_rectf(int x, int y, int w, int h)

Color a rectangle that exactly fills the given bounding box.

void fl_rectf(int x, int y, int w, int h, Fl_Color)

Color a rectangle with "exactly" the passed r,g,b color. On screens with less than 24 bits of color this is done by drawing a solid-colored block using fl_draw_image() so that dithering is produced.

void fl_rect(int x, int y, int w, int h)

Draw a line inside this bounding box (currently correct only for 0-thickness lines).

void fl_line(int x, int y, int x1, int y1)

Draw a straight line between the two points.

void fl_pie(int x, int y, int w, int h, double start, double end, int what=FL_PIE);

These functions match the rather limited circle drawing code provided by X and WIN32. The advantage over using fl_arc is that they are faster because they often use the hardware, and they draw much nicer small circles, since the small sizes are often hard-coded bitmaps. Only the integer translation of the current transformation is obeyed on most systems.

The allowed types are:

Text

See Fl_Font for a description of what can be passed as a font. For most uses one of the built-in constant fonts like FL_HELVETICA can be used.

void fl_font(Fl_Font, unsigned size)

Set the current font. It is scaled so that the point size is size pixels tall. The font size is unaffected by the current transformation.

void fl_font(const char* name, unsigned size);
void fl_font(const char* name, int attributes, unsigned size);

Set the current font by name. Exactly what names work depend on your system, it is best to use fl_list_fonts to see what is provided. See fl_find_font for how the name and attributes are interpreted.

void fl_encoding(const char*);

The encoding determines how the bytes sent to fl_draw are turned into glyphs. Unlike most toolkits there are no errors, if you pick an encoding that does not exist for this font, you will get some default encoding (for instance the Symbol font always works without having to set the encoding). The only way to find out what encodings are going to work is to call Fl_Font::encodings().

Notice that only 8-bit encodings are supported. In the future we plan to support UTF-8 which is called "iso10646", at that time support for 8-bit encodings may be dropped anyway. Notice that fltk will never support "wide" encodings, as I consider them evil.

Currently the default is "iso8859-1"

Fl_Font fl_font()

Returns the current font.

unsigned fl_size()

Returns the current font size.

const char* fl_encoding();

Returns the current encoding.

int fl_height()

Returns the vertical size of the font according to the system. It is highly recommended that you use fl_size() instead for portability and because many X fonts return erroneous values for this.

int fl_descent()

Recommended distance above the bottom of a fl_height() tall box to draw the text at so it looks centered vertically in that box.

int fl_width(const char*)
int fl_width(const char*, int n)
int fl_width(uchar)

Return the pixel width of a nul-terminated string, a sequence of n characters, or a single character in the current font.

void fl_draw(const char*, int x, int y)
void fl_draw(const char*, int n, int x, int y)

Draw a nul-terminated string or an array of n characters starting at the given location.

void fl_draw(const char*, int x, int y, int w, int h, Fl_Flags)

Fancy string drawing function which is used to draw all the labels. The string is formatted and aligned inside the passed box. Handles '\t' and '\n', expands all other control characters to ^X, and aligns inside or against the edges of the box. See
Fl_Labeltype_::draw() for values for the flags. The value FL_ALIGN_INSIDE is ignored, as this function always prints inside the box.

void fl_measure(const char*, int &w, int &h)

Measure how wide and tall the string will be when printed by the fl_draw(...align) function. If the incoming w is non-zero it will wrap to that width.

Images

If you plan to draw the same image many times, you may want an Fl_Image subclass such as Fl_Bitmap, Fl_RGB_Image, or Fl_Pixmap and call draw() on them. The advantage of using the object is that FLTK will cache translated forms of the image (on X it uses a server pixmap) and thus redrawing is much faster. In addition, on current systems, Fl_Image is the only way to get transparency or to draw 1-bit bitmaps.

The advantage of drawing directly is that it is more intuitive, and it is faster if the image data changes more often than it is redrawn.

Currently the image is only affected by the integer portion of the current transformation. This may change in future versions!

void fl_draw_image(const uchar*, int X, int Y, int W, int H, int D = 3, int LD = 0)
void fl_draw_image_mono(const uchar*, int X, int Y, int W, int H, int D = 1, int LD = 0)

Draw an 8-bit per color RGB or luminance image. The pointer points at the "r" data of the top-left pixel. Data must be in r,g,b order. X,Y are where to put the top-left corner. W and H define the size of the image. D is the delta to add to the pointer between pixels, it may be any value greater or equal to 3, or it can be negative to flip the image horizontally. LD is the delta to add to the pointer between lines (if 0 is passed it uses W * D), and may be larger than W * D to crop data, or negative to flip the image vertically.

It is highly recommended that you put the following code before the first show() of any window in your program to get rid of the dithering if possible:

Gray scale (1-channel) images may be drawn. This is done if abs(D) is less than 3, or by calling fl_draw_image_mono(). Only one 8-bit sample is used for each pixel, and on screens with different numbers of bits for red, green, and blue only gray colors are used. Setting D greater than 1 will let you display one channel of a color image.

The X version does not support all possible visuals. If FLTK cannot draw the image in the current visual it will abort. FLTK supports any visual of 8 bits or less, and all common TrueColor visuals up to 32 bits.

typedef void (*fl_draw_image_cb)(void*, int x, int y, int w, uchar *)
void fl_draw_image(fl_draw_image_cb, void*, int X, int Y, int W, int H, int D = 3)
void fl_draw_image_mono(fl_draw_image_cb, void*, int X, int Y, int W, int H, int D = 1)

Call the passed function to provide each scan line of the image. This lets you generate the image as it is being drawn, or do arbitrary decompression of stored data (provided it can be decompressed to individual scan lines easily).

The callback is called with the void* user data pointer (this can be used to point at a structure of information about the image), and the x, y, and w of the scan line desired from the image. 0,0 is the upper-left corner (not X,Y ). A pointer to a buffer to put the data into is passed. You must copy w pixels from scanline y, starting at pixel x , to this buffer.

Due to cropping, less than the whole image may be requested. So x may be greater than zero, the first y may be greater than zero, and w may be less than W. The buffer is long enough to store the entire W * D pixels, this is for convienence with some decompression schemes where you must decompress the entire line at once: decompress it into the buffer, and then if x is not zero, copy the data over so the x'th pixel is at the start of the buffer.

You can assume the y's will be consecutive, except the first one may be greater than zero.

If D is 4 or more, you must fill in the unused bytes with zero.