www.sbin.org

Xlib Programming Manual (O'Reilly & Associates, Inc.)

Color

A typical X application allows the user to specify colors for the background and border of each of its windows, colors for the cursor, and foreground and background colors to be set in GCs for drawing text and graphics. More complex applications (such as Computer Aided Design (CAD) applications) might use color to distinguish physical or logical layers. Still more complex applications, such as in imaging, might use fine gradations of color to represent real-world data. Yet in discussing the background and border window attributes and how to set the foreground and background members of the GC, we have spoken only of pixel values.

How are these pixel values translated to colors? And how must an X client manage color if it is to run successfully on the wide variety of screen hardware available in the X environment?

Because X must support a wide variety of systems with differing screen hardware, the Xlib color-handling mechanisms are fairly complex. Even programmers who have previously written color graphics applications will find there are some new concepts to learn.

This chapter starts out by describing the different types of screens that an X application may run on and the mechanisms Xlib provides for determining the screen type. It then describes the simplest color-allocation mechanisms, which could be used by applications whose principal use of color is for decoration. It proceeds to discuss more complex color applications and concludes with a section on writing applications that will be portable across different types of color and monochrome screens.

Basic Color Terms and Concepts

You, no doubt, know that a color screen uses multiple bits per pixel (also referred to as multiple planes) to specify colors. A colormap is used to translate each pixel's value into the visible colors you see on the screen.

A colormap is no more than a lookup table stored in the server. Any given pixel value is used as an index into this table--for example, a pixel value of 16 will select the sixteenth element, or colorcell.

On the most common type of color system, each colorcell contains separate 16-bit intensity values for each of the three primary colors.

As shown in Figure 7-1, a pixel value uniquely identifies a particular colorcell. Each pixel value in the visible portions of a window is continuously read out of screen memory and looked up in the colormap. The RGB values in the specified colorcell control the intensity of the three primary colors and thus determine the color that is displayed at that point on the screen.

Pixel value to RGB mapping with the colormap on a color screen

 The range of colors possible on the screen is a function of the number of bits available in the colormap for RGB specification. If eight bits is available for each primary, then the range of possible colors is 256 3 (about 16 million colors).

However, the number of different colors that can be displayed on the screen at any one time is a function of the number of planes. A 4-plane system could index 2 4 colorcells (16 distinct colors); an 8-plane system could index 2 8 colorcells (256 distinct colors); and a 24-plane system could index 2 24 colorcells (over 16 million distinct colors).

A client attempting to use color does not specify a pixel value and the color to be put in that cell in order to draw in a given color. Instead, it requests access to a colorcell in a colormap (managed by the server) and is returned a pixel value. This is called allocating a color. When a client allocates a color, it asks the server, "Which colorcell can I use?" and the server responds by saying, "You can use the colorcell specified by this pixel value." There are three basic functions that allocate colors, which are described in detail and demonstrated in later sections in this chapter.

Color Naming and Specification

The Server-side Color Name Database

Because of differences in screen hardware, the same RGB values may generate quite different colors on different hardware. Therefore, server implementors were intended to change the RGB values corresponding to each color name to make sure that the appropriate color appears on their screen. This is called gamma correction. By using names from this database, you are more sure of getting a color close to the one you request. If the server implementor has not provided a gamma-corrected color database, there is no way a program can tell exactly what color is being displayed even when it knows the RGB values. This problem is solved with the X Color Management System, or Xcms, which was introduced in Release 5. Xcms provides a client-side color database, and supports device-independent color specification. In R5, color name strings are looked up first with the client-side color database, and if not found then on the server database.

It is also important to note that the color names are not specified by the X11 protocol or Xlib. Therefore, server implementors may change them, but more often, they will simply add to the list. (Note that some servers allow users to customize this file. For more information, see Volume Three, X Window System User's Guide.)

Table 7-1 shows some of the color names and corresponding RGB values in the default color database. The complete R4 database is extensive and is shown in Appendix D, The Server-side Color Database, of Volume Two, Xlib Reference Manual. The text version of this database in the standard distribution on a UNIX-based system is in the file /usr/lib/X11/rgb.txt. The server reads a compiled version of it.

The color names in the color database are strings in which each character uses the ISO Latin-1 encoding. The ISO (International Standards Organization) Latin-1 encoding is used by virtually all workstations manufacturers. What this means is that the first 127 character codes correspond to 7-bit ASCII and are the normal English characters that appear on U.S. keyboards. But ISO characters are 8-bit, and the characters from 128 to 255 are used for characters with accents and other variations, necessary for other Western languages.

Server vendors should be able to supply a color database file for each foreign language. The RGB values would be the same, but the names would be different. In the English file, the entry for green is encoded with the ISO character codes 103 (g), 114 (r), 101 (e), 101 (e), 110 (n). In German, the same entry would be for grün, encoded with the ISO codes 103 (g), 114 (r), 252 (ü), 110 (n). In a workstation configured for German, there will be an easy way to type ü.

Note that keysyms also use the ISO Latin-1 standard, as shown in Chapter 8, "Events."
 

Table 7-1 : Sample from the Server-side Color Database*

English Words	Red	Green	Blue
aquamarine	112	219	147
black	0	0	0
blue	0	0	255
blue violet	159	95	159
brown	165	42	42
cadet blue	95	159	159
coral	255	127	0
cornflower blue	66	66	111
cyan	0	255	255
light gray	168	168	168
light grey	168	168	168
light steel blue	143	143	188
lime green	50	204	50
magenta	255	0	255
maroon	142	35	107
medium aquamarine	50	204	153

 

Also defined are the color names "gray0" through "gray100", spelled with an "e" or an "a". "gray0" is black and "gray100" is white. See Appendix D, The Server-side Color Database, of Volume Two, Xlib Reference Manual, for a listing of the complete sample database.

Xcms Color Specification

A device-dependent RGB value is represented as follows:

RGB:<red>/<green>/<blue>

X11R5 supports an additional device-dependent color space, called RGBi, in which each red, green, and blue integer value is replaced with a floating-point intensity between 0.0 and 1.0. In this model, the range of possible color values are simply mapped onto the real numbers between zero and one. So, for example, 0.5 always represents half intensity of a color. Note that these values represent the physical intensity of a color, which is not linearly proportional to the perceptual intensity of that color. A color specification for RGBi has the following form:

RGBi:<red>/<green>/<blue>

Device-independent color specifications follow the same syntax--a color space name followed by a colon and slash-separated color space values. The following forms are recognized:

CIEXYZ:<X>/<Y>/<Z> 
CIEuvY:<u>/<v>/<Y> 
CIExyY:<x>/<y>/<Y> 
CIELab:<L>/<a>/<b> 
CIELuv:<L>/<u>/<v> 
TekHVC:<H>/<V>/<C>

Example 7-1 shows this new style of color specification used in a resource file. Notice that color space names are case-insensitive.

Specifying device-independent colors from a resource file

*Background: TekHVC:72.0/50.0/44.0
*Command.background: CIELab:75.0/.38/.71
*quit_button.background: rgbi:1.0/0.0/0.0

The Client-side Color Name Database

Note that while X11R5 supports such a color database, the MIT release does not provide one, other than as an example to system administrators or users who want to define one of their own. The client-side color database should be thought of as a place for site-specific customizations, and useful, if non-standard, shortcuts for naming colors in user resource files. In particular, since the contents of the database are not standardized, application defaults files should not rely on any particular colors to be in the database.

X clients (on most UNIX systems) look for the client-side database in the file /usr/lib/X11/Xcms.txt by default, but the MIT sample implementation allows a different file to be specified with the XCMSDB environment variable. The format of the database is implementation-dependent. Example 7-2 shows an example database in the format supported by the MIT distribution.

Example entries from a client color database

XCMS_COLORDB_START 0.1
device red      RGBi:1.0/0/0
device blue     RGB:00/00/ff
navy blue       CIEXYZ:0.0671/0.0337/0.3130
gray0           CIELab:0.0/0.0/0.0
gray50          CIELuv:50.0/0.0/0.0
grey100         TekHVC:0.0/100.0/0.0
rouge           red
roja            rouge
XCMS_COLORDB_END

When the functions XAllocNamedColor(), XLookupColor(), XParseColor(), XStoreNamedColor(), or their device-independent Xcms analogs are passed a color string, they first attempt to parse it as a new-style specification for one of the supported color spaces. If this fails, they attempt to look up the color in the client-side color name database. If both approaches fail, they fall back on the pre-X11R5 behavior and attempt to parse the string in the old-style numeric format or pass the string to the X server to be looked up in the server database. Because the new X11R5 formats are supported by the pre-X11R5 Xlib functions, all X Toolkit widgets and type converters will work correctly with device-independent color specifications without change.

Hexadecimal Color Specification

The hexadecimal form of color specification is necessary in R4 for the user to be able to specify an exact color, not just the rough approximation allowed by an string name. The hexadecimal specification must be in one of the following formats:

#RGB            (4 bits each of red, green, and blue)
#RRGGBB         (8 bits each of red, green, and blue)
#RRRGGGBBB      (12 bits each of red, green, and blue)
#RRRRGGGGBBBB   (16 bits each of red, green, and blue)

Use of hexadecimal color specifications does not preclude colorcell sharing, since the user could specify the same hexadecimal value for the color for two or more clients. However, it probably tends to make sharing less likely, since a window manager might allocate all the colors in the color database as read-only cells, and then any client that uses hexadecimal specifications will probably be allocating a separate cell instead of sharing.

Differences in Display Hardware

Mid-range Color Displays

Mid-range color screens usually have only one hardware colormap. In other words, the pixel values in all the windows on the screen are mapped to colors using the same colormap. On most of these systems, however, the color in any colorcell in the hardware colormap can be individually changed, and therefore, the entire colormap can be replaced with a new set of values. X provides the concept of the virtual colormap, so that more than one set of colorcells can be maintained, even though only one of them can be in use at a time. Virtual colormaps are swapped in and out of the hardware colormap by the window manager. This makes it possible for an application that has special color needs to create its own virtual colormap, which the window manager will load into the hardware colormap when that application is in use. However, since only one hardware colormap is available and all applications share it, when any one application creates a new virtual colormap and the window manager installs it, all other applications will screen in false colors, since the pixel values they use now point to cells in the other client's colormap. This is acceptable, since the window manager always installs the correct colormap for the application in use, but it is obviously not ideal. On high performance systems, described below, this problem is solved by having multiple hardware colormaps.

Monochrome and Gray Scale

Gray-scale screens are sometimes used for publishing applications, since pixels made up of a single phosphor are smaller than those made up of three phosphors and the resolution is, therefore, better. As shown in Figure 7-2, a gray-scale screen works by looking up the intensity of the pixel in the colormap, which, for this screen type, contains only a single value. This controls the intensity of a single electron beam. Gray scale can be simulated on a color screen by making the red, green, and blue values equal in a given colorcell to determine the brightness of gray pixels on the screen.

Pixel value to RGB mapping -- gray scale and monochrome screens

 A gray-scale screen might have a read-only colormap, so that the gray levels in each cell could not be changed. A monochrome screen is an example of this type; it is a single-plane screen with a two-element read-only colormap.

High Performance Color Displays

The problem with having so many planes is that a colormap of the style used in mid-range color screens would be impossibly large: it would contain over 16 million entries. Instead, the available bits per pixel are broken down into three separate colormap indices, one for each primary color, as shown in Figure 7-3. This approach still allows the full range of colors to be generated but makes the job of loading the colormap much more manageable. This scheme requires three primary colormaps of only 256 entries each to specify all 16 million colors on a 24-plane system.

Pixel value to RGB mapping -- high performance color screens

 In high performance screens, having a read-only colormap makes just as much sense as having it read/write, because nearly every color imaginable can be simultaneously available. With a read-only colormap, there is a fixed relationship between the pixel values used to select a color and the actual RGB values generated. This makes possible applications that want to calculate pixel values directly instead of having to calculate colors and then determine which pixel value represents that color, as is necessary when the colormap is read/write.

In reality, most screens in this class let you use the color resources in either fashion, using virtual colormaps. There can be one read-only virtual colormap and one read/write virtual colormap. However, unlike on mid-range color screen hardware, most high performance color systems have multiple hardware colormaps, so that both virtual colormaps can be installed and used at the same time. In fact, on many of these systems, each window can have its own virtual colormap installed in the hardware at the same time.

How X Describes Color Support with Visuals

Most windows inherit their parent's visual, and windows will often share the root window's visual, which is known as the default visual. The default visual describes, naturally, the default colormap. If you create all your windows with XCreateSimpleWindow(), you will be using the default visual and colormap.

The Visual structure is intended to be opaque; programs are not supposed to access its contents. This is so that Xlib implementors can change the structure without breaking existing clients. The procedure used to avoid accessing its members is not all that cumbersome but is just beginning to come into use by application writers. Up to this point, most programmers have broken this rule. We will show you only the correct method here, since it adds only a few lines to the application.

Even more existing applications have avoided visuals altogether and used only the DefaultDepth() or DisplayPlanes() macros to attempt to determine whether the screen is monochrome or color. However, this does not work in general, because it does not distinguish between gray-scale screens and color screens (both have more than one plane). The only way to make this distinction is to get information about visuals.

Remember that a visual is only one way to use color on a particular screen. There may be a list of supported visuals on a screen, with each visual describing a different depth and writeability of the colormap. On a color system, there may be both monochrome and color visuals available.

The correct method to get information about the visuals supported on a particular screen is to use XMatchVisualInfo() or XGetVisualInfo(). These functions return XVisualInfo structures that contain information about the available visuals and are public so their fields can be safely accessed.

The class member of XVisualInfo contains a constant specifying one of six different visual classes,  corresponding to the basic ways of using a screen: DirectColor, GrayScale, PseudoColor, StaticColor, StaticGray, or TrueColor.

As summarized in Table 7-2, the visual classes distinguish between color or monochrome, whether the colormap is read/write or read-only, and whether a pixel value provides a single index to the colormap or is decomposed into separate indices for red, green, and blue values.
 

Table 7-2 : Comparison of Visual Classes

Colormap Type	Read/Write	Read-only
Monochrome/Gray	GrayScale	StaticGray
Single Index for RG&B	PseudoColor	StaticColor
Decomposed Index for RG&B	DirectColor	TrueColor

 

There may be more than one way of using color on a particular screen, and therefore, there may be more than one supported visual. This is usually true of high-end workstations. There are ways to search through the available visuals to select the one that most closely meets the needs of your application, as will be described later. Several visuals of the same class may be provided but at different depths. On high performance screens, it is possible to create the colormap as read/write or as read-only. Both methods have certain advantages and would be used for different applications. There would be a separate visual for each of these ways of using the screen hardware. One of these visuals would be TrueColor class and the other DirectColor class. Some 24-plane screens allow the screen to be treated as two separate 12-plane PseudoColor visuals. (This allows for "double-buffering," a technique useful for animation, or for storing distance data to simplify hidden line and plane calculations in 3-D applications.) In fact, on some advanced workstations, you can use a different visual in each window.

Figure 7-4 schematically represents the visual classes that can theoretically be supported by each type of screen hardware. A screen that supports the DirectColor class can theoretically support any of the six visual classes. A screen that supports the PseudoColor visual class can support GrayScale, PseudoColor, StaticColor, or StaticGray visual classes. A screen that supports the GrayScale visual class can also support StaticGray visual classes. The three types of screen with read-only colormaps can only support visuals of their own class. But remember that just because a certain visual class can theoretically be supported by a certain screen hardware does not mean that the server implementors will decide to support that class.

Hierarchy of visual classes

Shareability vs. Changeability

One advantage of immutable colormaps is that all the cells are read-only and can be shared between clients, so all the cells are available to every client. Immutable colormaps also make it possible to calculate pixel values from the colors desired without querying the server, since the mapping between pixel values and colors is predictable. This technique is necessary for smooth shading and 3-D rendering algorithms. As you will see, this is usually not possible with changeable colormaps. The disadvantages of immutable colormaps are that there may not be the exact color you desire (if there are a small number of planes) and you cannot allocate read/write cells, so you cannot change a colorcell to change the color of existing pixels on the screen. To change a color, you have to redraw the graphics with a new pixel value.

In general, the advantage of changeable colormaps is that you can have both private read/write cells and shareable read-only cells. That is why PseudoColor and DirectColor are the most useful visuals, when a screen supports them. PseudoColor and DirectColor allow you to decide whether your client really needs read/write cells or whether it can use read-only cells. Read-only usage is preferred, since these cells can be shared by all clients, which means that the colormap is less likely to run out of free cells.

Try not to confuse the writeability of colormaps with the writeability of colorcells. A colorcell in a read/write colormap can be allocated read/write or read-only. A colorcell in a read-only colormap can only be allocated read-only. A changeable colormap could be made entirely read-only if the window manager or any other client allocates all available colorcells read-only.

The advantages of read/write colorcells, available only in changeable colormaps, are that your program can select exactly the color you want (as long as it is physically possible on the screen) and you can change the color at will, which instantly changes the visible color of everything drawn with that pixel value if the colormap is currently installed. Although any other client can also change the values in a read/write cell, it is a convention that only the client that allocated the cell should change its contents. You own that pixel value. Since most clients cannot be satisfied with having no control over their displayed colors, this pixel value is not shareable. That means that if several clients that use read/write colorcells are running, all the colorcells might be used. Then some client will be forced to create its own colormap, with the negative consequences described in Section 7.10, "Creating and Installing Colormaps."

Allocating Shared Colors

The returned pixel value can be used to set the background_pixel or border_pixel attribute of a window or to set the foreground or background member of a GC, which are used by drawing requests. (See Chapter 4, "Window Attributes," and Chapter 5, "The Graphics Context," for more information.)

Read-only colorcells can be allocated with the following routines:

XAllocColor()

Returns the index of the colorcell (a pixel value) that contains the RGB values that are requested or that contains the closest RGB values physically possible on the screen. XcmsAllocColor() is the same except that colors are specified using Xcms syntax.

XAllocNamedColor()

Returns the index of the colorcell that contains the RGB values associated with a specified color name from the string color name database or the closest RGB values physically possible on the screen. In R5, XAllocNamedColor() accepts strings in Xcms syntax. XcmsAllocNamedColor() is almost the same except with a different style of arguments. XcmsAllocNamedColor() also is capable of returning a symbol describing the format found in the specified string.

XParseColor() parses a color name string or a hexadecimal color specification string and returns RGB values. It can be used with XAllocColor() or the routines that allocate read/write cells, which will be described later. For color names, it gets the RGB values from the server's color database just like XAllocNamedColor(). You may have noticed that XAllocNamedColor() is very similar to the combination of XParseColor() and XAllocColor(). The difference is slight: XAllocNamedColor() can interpret color names but not hexadecimal specifications--but hexadecimal specifications are rarely made by users anyway. The two-routine combination is more often used because it allows you to separately report errors in parsing the color specified and allocating the colorcell.

Using XQueryColor() and XQueryColors() you can find out what RGB values are in each colorcell. But there is no way to determine whether a given cell is read-only or read/write. The only way to tell how many cells are currently unallocated is to allocate N colors using XAllocColorCells(), using the maximum possible N initially, then reduce N until it succeeds (not recommended). A binary search is a faster way to find N.

A request to allocate a color may fail because there are no free colormap cells and, for read-only colorcells, because no existing colorcell contains the closest color possible on the hardware to the exact color requested. Applications must allocate colors by trial and error. The routines that allocate colorcells all have Status return values. If the call to allocate colorcells returns False, the client may modify the arguments and try again. If repeated attempts fail, the client can settle with BlackPixel and WhitePixel() or, if these colors are inadequate, create a new virtual colormap. An application with picky color needs that cannot be satisfied can simply report to the user that its color needs cannot be met and exit.

Note that XAllocColor() works somewhat differently on dynamic visuals (such as PseudoColor) than it does on static visuals (like StaticColor). On dynamic visuals, it fails if it is unable to allocate the exact RGB values requested (i.e., if there are no free cells and no cells already allocated with the exact RGB values requested). On static visuals, it returns the cell with the closest RGB values. The algorithm used to determine "closest" is server-dependent. The moral of this is that a program must be prepared for XAllocColor() to fail.

The XColor Structure

The XColor structure is shown in Example 7-3. The information it contains closely matches the information in each cell of the colormap.

The XColor structure

typedef struct {
   unsigned long pixel;               /* Pixel value */
   unsigned short red, green, blue;   /* RGB values */
   char flags;                        /* DoRed,  DoGreen, and/or
                                       * DoBlue */
   char pad;                          /* Unused; pads structure
                                       * to even word boundary */
} XColor;

The red, green, and blue members are 16-bit values. Full brightness in a color is a value of 65535, half brightness is 32767, and off is 0. (The server automatically scales these values if the hardware colormap includes fewer bits for RGB values.)

The flags member of the XColor structure is a bitwise OR of the symbols DoRed, DoGreen, and DoBlue. These flags are used to specify which of the red, green, and blue values should be read while changing the RGB values in a read/write colorcell. How these are used is demonstrated in Example 7-4.

Code to Allocate Read-only Colors

This routine uses XMatchVisualInfo() to determine whether color is supported on the screen. If any of the four color visual classes are supported, it proceeds to attempt to allocate read-only colors. Whenever anything fails or if color is not supported, the routine uses black and white. For some applications, this could be modified to allocate levels of gray on GrayScale visual class screens.

The code for all the examples in this chapter is in the example source in the directory /basicwin/color/. This example is called basic.ro.

Allocating read-only colorcells

#include <X11/Xlib.h>
#include <X11/Xutil.h>
#include <X11/Xos.h>
#include <stdio.h>
extern Display *display;
extern int screen_num;
extern Screen *screen_ptr;
extern unsigned long foreground_pixel, background_pixel,
      border_pixel;
extern char *progname;
#define MAX_COLORS 3
/* This is just so we can print the visual class intelligibly */
static char *visual_class[] = {
   "StaticGray",
   "GrayScale",
   "StaticColor",
   "PseudoColor",
   "TrueColor",
   "DirectColor"
};
get_colors()
{
   int default_depth;
   Visual *default_visual;
   static char *name[] = {"Red", "Yellow", "Green"};
   XColor exact_def;
   Colormap default_cmap;
   int ncolors = 0;
   int colors[MAX_COLORS];
   int i = 5;
   XVisualInfo visual_info;

   /* Try to allocate colors for PseudoColor, TrueColor,
    * DirectColor, and StaticColor; use black and white
    * for StaticGray and GrayScale */
   default_depth = DefaultDepth(display, screen_num);
   default_visual = DefaultVisual(display, screen_num);
   default_cmap   = DefaultColormap(display, screen_num);
   if (default_depth == 1) {
      /* Must be StaticGray, use black and white */
      border_pixel = BlackPixel(display, screen_num);
      background_pixel = WhitePixel(display, screen_num);
      foreground_pixel = BlackPixel(display, screen_num);
      return(0);
   }
   while (!XMatchVisualInfo(display, screen_num, default_depth,
         /* visual class */i--, &visual_info))
      ;
   printf("%s: found a %s class visual at default depth.\n",
         progname, visual_class[++i]);

   if (i < StaticColor) { /* Color visual classes are 2 to 5 */
      /* No color visual available at default depth;
       * some applications might call XMatchVisualInfo
       * here to try for a GrayScale visual if they
       * can use gray to advantage, before giving up
       * and using black and white */
      border_pixel = BlackPixel(display, screen_num);
      background_pixel = WhitePixel(display, screen_num);
      foreground_pixel = BlackPixel(display, screen_num);
      return(0);
   }
   /* Otherwise, got a color visual at default depth */
   /* The visual we found is not necessarily the default
    * visual, and therefore it is not necessarily the one
    * we used to create our window; however, we now know
    * for sure that color is supported, so the following
    * code will work (or fail in a controlled way) */
   /* Let's check just out of curiosity: */
   if (visual_info.visual != default_visual)
   {
      printf("%s: %s class visual at default depth\n",
            progname, visual_class[i]);
       printf("is not default visual! Continuing anyway...\n");
   }
   for (i = 0; i < MAX_COLORS; i++) {
      printf("allocating %s\n", name[i]);
      if (!XParseColor (display, default_cmap, name[i],
            &exact_def)) {
         fprintf(stderr, "%s: color name %s not in database",
               progname, name[i]);
         exit(0);
      }
      printf("The RGB values from the database are %d, %d, %d\n",
            exact_def.red, exact_def.green, exact_def.blue);
         if (!XAllocColor(display, default_cmap, &exact_def)) {
         fprintf(stderr, "%s: can't allocate color:\n",
               progname);
         fprintf(stderr, "All colorcells allocated and\n");
         fprintf(stderr, "no matching cell found.\n");
         exit(0);
      }
      printf("The RGB values actually allocated are %d, %d, %d\n",
            exact_def.red, exact_def.green,
            exact_def.blue);
      colors[i] = exact_def.pixel;
      ncolors++;
   }
   printf("%s: allocated %d read-only color cells\n",
         progname, ncolors);
   border_pixel = colors[0];
   background_pixel = colors[1];
   foreground_pixel = colors[2];
   return(1);
}

As noted in the code, the visual found might not necessarily be the default visual. This does not always matter, because if any color visual is available, it is a good bet that the default visual is also color, and so colors can be allocated without doing any further research. With XMatchVisualInfo(), it is difficult to develop an algorithm that is guaranteed to find the default visual. This is much easier with XGetVisualInfo(), which returns a list of available visual structures that match a set of criteria you specify. If you pass no criteria, it simply returns the entire list of available visuals. You can then search through the list matching the visual member of the XVisualInfo structures to the default visual. This will be demonstrated in Section 7.6, "Getting Complete Visual Information."

The XParseColor() call specifies a color name, and the RGB values corresponding to that name are returned from the color database in the passed XColor structure. This structure is then passed to XAllocColor(), and the pixel value allocated is returned in the pixel field of the structure.

The same calls would be used to parse a hexadecimal color string. Pink could be specified in the call to XParseColor() as "#bc8f8f" instead of "pink". But, as we have said before, color names are preferred, because there is a better chance that they will specify a color already allocated or later to be allocated by another client.

It is also possible to specify the desired RGB values explicitly. This is good for default colors because it saves a call to XParseColor(), but on the other hand, you might not get a consistent color on all systems because you are bypassing the gamma correction implemented through the color database. Simply declare an XColor structure and set its red, green, and blue members to the desired RGB values. Of course, these values can be specified as integers, hexadecimal values, or any other way that the C language allows. Then pass this structure to XAllocColor(). But remember, as we have said, it is better to use color names when allocating read-only colorcells than to use any of these explicit RGB values.

Highlighting in Two Colors

When drawing in BlackPixel() and WhitePixel() on a color system, the color drawn by the GXxor operation is random if BlackPixel() or WhitePixel() are used for the foreground pixel value in the GC. This is because there is no restriction on which pixel value BlackPixel() and WhitePixel() can be on a server--they are not necessarily 1 and 0 and not necessarily different by just one bit. For example, the pixel value drawn if the foreground pixel value in the GC is BlackPixel() and the pixel value on the screen is WhitePixel() is BlackPixel() XOR WhitePixel(), which, unless BlackPixel() and WhitePixel() are different by only one bit, is a third pixel value not allocated by this client. The colorcell identified by this pixel value might contain black, in which case the operation would not change the screen.

The solution to this problem, which works on monochrome and color systems, is to set the foreground pixel value in the GC used in drawing with GXxor to the exclusive OR of BlackPixel() and WhitePixel() or by setting the logical function to GXinvert and using a plane mask which is the exclusive OR of WhitePixel() and BlackPixel(). All applications that highlight graphics drawn in BlackPixel() and WhitePixel() on a color system should use one of these two methods. The following example illustrates how this works using two arbitrarily chosen pixel values (which could be BlackPixel() and WhitePixel() or could be any two colors).

Let's assume that we draw in two pixel values, which we will call color1 and color2. The pixel values for these could be:

color1 = 11111111111111110000000000000000
color2 = 00000000111111111111111100000000

color1 XOR color2 = 11111111000000001111111100000000

foreground =               11111111000000001111111100000000
existing pixel (color1) =  11111111111111110000000000000000
resulting pixel (color2) = 00000000111111111111111100000000

Choosing Default Colors

Follow these guidelines for your application's default colors:

• Use string color names for read-only colorcells if possible, since this maximizes the chance of sharing cells.
• Use colors with large contributions from two or all three primary colors--they light the screen more brightly.
• Avoid shades of pure blue--the human eye is relatively insensitive to and unable to focus on images made of pure blue light. Mix blue shades with white (white contains equal parts of all three primary colors).
• Remember that some users are color blind. Do not use the same intensity of green and red for "safe" and "danger"--use colors with differing intensity.

Allocating Private Colors

• The application draws something whose color must be changed dynamically without redrawing it. For example, in a color mixing program, the palette must be drawn in colors that change frequently. If this were done with read-only colors, cells would have to be allocated and freed frequently and the palette area redrawn with each new color. However, with read/write colorcells, the steps of allocation and color setting are separate, so that the color of an already allocated cell can be changed at will. Anything drawn using the pixel value of this colorcell will change color immediately when the RGB values in the colorcell are changed.
• The application needs to overlay graphics on top of other graphics in such a way that the overlayed graphics can be erased without disturbing the underlying graphics. For example, in a Computer Aided Design (CAD) package for chip design, it is often useful to overlay the various layers of a chip in different colors on the screen. When one of the layers is removed, you want to avoid having to redraw all the underlying layers. How to do this by allocating read/write cells will be described.
• The system has a huge colormap, and the application needs to set a large number of colorcells. The calls for manipulating read/write colorcells allow you to manipulate multiple cells per call, whereas with read-only cells, you are limited to one cell per call.

XAllocColorCells() allocates read/write colorcells. At its simplest, it allows you to allocate read/write cells so you can change the RGB values dynamically.

But to simply allocate just a few cells, you set the ncolors argument to the number of colorcells desired and nplanes to 0, and all the pixel values you need will be returned in the pixels array. The real reason for the nplanes and plane_masks arguments will become clear in Section 7.5.2, "Allocating Read/Write Colorcells for Overlays." The RGB values of the allocated cells are set with XStoreColor(), XStoreColors(), or XStoreNamedColor().

XAllocColorPlanes(), on the other hand, is only used when you want to be able to vary a primary color component of graphics already drawn without redrawing them. It allocates read/write cells, so that a preset number of bits are reserved for each primary color. Primarily for DirectColor, it also allows you to simulate a small DirectColor colormap on a PseudoColor visual but uses up colorcells quickly. It treats the colormap as three separate lookup tables, allocating ncolors*2nreds entries in the red lookup table, ncolors*2ngreens entries in the green lookup table, and ncolors*2nblues entries in the blue lookup table.

The following routines are used to actually store colors into read/write colorcells once they are allocated:

XStoreColor()

Changes the read/write colormap cell corresponding to the specified pixel value to the hardware color that most closely matches the RGB values specified.  The flags DoRed, DoGreen, and DoBlue in the XColor structure indicate which primary colors in the cell are to be changed. XcmsStoreColor() is similar but allows you to specify the color string in Xcms syntax.

XStoreColors()

Like XStoreColor(), except it does multiple cells per call. Changes the read/write colormap cell corresponding to the specified pixel value to the hardware color that most closely matches the RGB values specified. The flags DoRed, DoGreen, and DoBlue in each XColor structure indicate which primary colors in each cell are to be changed. XcmsStoreColors() is similar but allows you to specify color strings in Xcms syntax.

XStoreNamedColor()

Performs the same function as StoreColor(), except that it stores the RGB values associated with a string color name in the RGB database. This call would be useful for loading a private colormap with each of the default named colors. No Xcms equivalent, since this function accepts Xcms syntax (in R5 and later).

Allocating Read/Write Colorcells for Dynamic Colors

Allocating read/write colorcells for dynamic colors

#include <X11/Xlib.h>
#include <X11/Xutil.h>
#include <X11/Xos.h>
#include <stdio.h>
extern Display *display;
extern int screen_num;
extern unsigned long foreground_pixel, background_pixel,
      border_pixel;
#define MAX_COLORS 3
get_colors()
{
   int default_depth;
   Visual *default_visual;
   static char *name[] = {"Red", "Yellow", "Green"};
   XColor exact_defs[MAX_COLORS];
   Colormap default_cmap;
   int ncolors = MAX_COLORS;
   int plane_masks[1];
   int colors[MAX_COLORS];
   int i;
   XVisualInfo visual_info;
   int class;
   class = PseudoColor;
   default_depth = DefaultDepth(display, screen_num);
   default_visual = DefaultVisual(display, screen_num);
   default_cmap   = DefaultColormap(display, screen_num);
   if (default_depth == 1) {
      /* Must be StaticGray, use black and white */
      border_pixel = BlackPixel(display, screen_num);
      background_pixel = WhitePixel(display, screen_num);
      foreground_pixel = BlackPixel(display, screen_num);
      return(0);
   }
   if (!XMatchVisualInfo(display, screen_num, default_depth,
         PseudoColor, &visual_info)) {
      if (!XMatchVisualInfo(display, screen_num, default_depth,
            DirectColor, &visual_info)) {
         /* No PseudoColor visual available at default_depth;
           * some applications might try for a GrayScale
           * visual here if they can use gray to advantage,
           * before giving up and using black and white */
         border_pixel = BlackPixel(display, screen_num);
         background_pixel = WhitePixel(display, screen_num);
         foreground_pixel = BlackPixel(display, screen_num);
         return(0);
      }
   }
   /* Got PseudoColor or DirectColor visual at default_depth */
   /* The visual we found is not necessarily the default
    * visual, and therefore it is not necessarily the one
    * we used to create our window; however, we now know
    * for sure that color is supported, so the following
    * code will work (or fail in a controlled way) */
   /* Allocate as many cells as we can */
   ncolors = MAX_COLORS;
   while (1) {
         if (XAllocColorCells (display, default_cmap, False,
            plane_masks, /* nplanes */0, colors, ncolors))
         break;
   ncolors--;
   if (ncolors == 0)
      fprintf(stderr, "basic: couldn't allocate read/write \
            colors\n");
      exit(0);
   }
   printf("basic: allocated %d read/write color cells\n", ncolors);
   for (i = 0; i < ncolors; i++) {
      if (!XParseColor (display, default_cmap, name[i],
            &exact_defs[i])) {
         fprintf(stderr, "basic: color name %s not in database",
               name[i]);
         exit(0);
      }
      /* Set pixel value in struct to the allocated one */
      exact_defs[i].pixel = colors[i];
            exact_defs[i].flags = DoRed | DoGreen | DoBlue;
   }
   /* This sets the color of read/write cell */
   XStoreColors (display, default_cmap, exact_defs, ncolors);
   border_pixel = colors[0];
   background_pixel = colors[1];
   foreground_pixel = colors[2];
}

Main of basic.rw -- changing colors of dynamic colorcells

  .
  .
  .
void main(argc, argv)
int argc;
char **argv;
{
     .
     .
     .
   XColor color;
   unsigned short red, green, blue;
     .
     .
     .
   /* Open display, etc. */
   color.pixel = foreground_pixel;
   XQueryColor(display, DefaultColormap(display, screen_num),
         &color);
   printf("red is %d, green is %d, blue is %d\n", color.red,
         color.green, color.blue);
   while (1)  {
      XNextEvent(display, &report);
      switch  (report.type) {
        .
        .
        .
      case ButtonPress:
         color.red += 5000;
         color.green -= 5000;
         color.blue += 3000;
         printf("red is %d, green is %d, blue is %d\n",
               color.red,
         color.green, color.blue);
         XStoreColor(display, DefaultColormap(display,
               screen_num), &color);
         break;
        .
        .
        .
   }
}

Allocating Read/Write Colorcells for Overlays

The trick that allows drawing without destroying what is already drawn relies on the fact that we can draw in one plane of the drawable, changing the pixel values and therefore the color, without changing any other plane. This is possible using the plane_mask component of the GC. It is these other planes that contain the information about the drawing that was already there. The disadvantage of this approach is that we have to allocate more colorcells than we would normally need. Some of the colorcells will need to be loaded with duplicate RGB values. Because of this waste of colorcells, this technique should be used only when the graphics being preserved are slow for the client or the server to redraw.

To illustrate this trick, we are going to draw in one color (the foreground in the GC), set the background_pixel attribute of the window to a second color, and then draw something temporary over the top with a third color.  To do this, we need to allocate four colorcells with XAllocColorCells(). The pixel values allocated will look something like this:

Color             Important Bits                            Remaining Bits                         
foreground:    ----0--0------------------------  all other bits don't matter
background:    ----0--1------------------------  all other bits don't matter
highlight1:    ----1--0------------------------  all other bits don't matter
highlight2:    ----1--1------------------------  all other bits don't matter

XAllocColorCells() does not return these four pixel values directly. Instead it returns the arrays colors and plane_masks that are more convenient for actually using the overlays than a single array of pixel values. (Each of these arrays has the number of members that was specified in the ncolors and nplanes arguments.) Both arrays consist of unsigned long values like pixel values. One array contains the plane masks of the overlay planes, and the other contains the pixel values that can be used for drawing independent of the overlay planes. Here are the values returned in each array after we call XAllocColorCells() with ncolors = 2 and nplanes = 1. These values are then used to generate the pixel values shown above.

Array Members       Important Bits                       Remaining Bits
colors[0]      =  ----0--0------------------------  other bits don't matter
colors[1]      =  ----0--1------------------------
plane_masks[0] =  ----1---------------------------  all other bits 0

Note that highlight2, generated with (colors[1] | plane_masks[0]), can be useful. As mentioned earlier, highlight2 can be used interchangeably with highlight1, as long as the plane_mask in the GC is set to plane_masks[0]. But highlight2 has another use. With a GC that does not have its plane_mask set to plane_masks[0] (the GC used for drawing with the foreground or background), this fourth pixel value can be used for drawing in the highlighting color while wiping out the underlying graphics, so that when the highlight is removed, the background color appears regardless of the contents of the drawable before the highlighting.

We have been hinting at the fact that this overlay technique can be used with more than two colors and more than one plane. ncolors specifies the number of colors than can be drawn and preserved while drawing in the overlays. nplanes specifies how many separate one-color overlays you may have or how many bits of color are available in a single overlay. The pixel values in the colors array are the ones that will be preserved through overlays. By ORing together each colors with any combination of plane_masks, you get the pixel values that are used for drawing the overlays. Note, however, that the plane mask of the GC used for the overlaying must be the OR of the same combination of members of the plane_masks array as were used to generate the pixel value.

The total number of pixel values (colorcells) allocated by XAllocColorCells() is ncolors*2nplanes . Note that the more planes you try to allocate, the less likely this request is to succeed, particularly on PseudoColor visuals. Therefore, if you are trying for multiple overlays or one multicolor overlay, this will probably work reliably only on DirectColor visuals, so make sure you have a backup plan for more common systems. In most cases, the underlying graphics can be redrawn if the overlays that would preserve them cannot be allocated. It is also possible to use backing store (which can save selective planes) or to manage your own off-screen pixmaps for use in fast redrawing of complicated graphics.

XAllocColorCells() takes a contig argument that specifies whether the planes returned in plane_masks must be contiguous. The contig argument is normally set to False, specifying that the allocated planes need not be contiguous, because then the chances of success of the XAllocColorCells() call are greater. There are more likely to be a number of noncontiguous planes available than the same number of contiguous planes. The contig argument may have to be set to True for imaging applications that want to be able to perform mathematical operations on the pixel values. It is easier to perform operations by shifting bits with contiguous planes than to achieve the same effect with random planes.

Each plane mask has one bit for GrayScale and PseudoColor or three bits for DirectColor or TrueColor, and none of the masks have bits in common.

Example 7-4 demonstrates allocating the read/write cells for a single overlay plane. It implements the overlay scheme described above. If this overlay plan fails, it allocates three colors so that a highlight can still be implemented even though the underlying graphics will have to be redrawn. If the color allocation fails completely, it uses black and white, which can be highlighted using the GXxor logical function to invert the color, as described in Section 7.4.3, "Highlighting in Two Colors."

Using XAllocColorCells() to allocate read/write colorcells for overlay plane

#include <X11/Xlib.h>
#include <X11/Xutil.h>
#include <X11/Xos.h>
#include <stdio.h>
extern Display *display;
extern int screen_num;
extern unsigned long foreground, background_pixel, overlay_pixel_1,
      overlay_pixel_2;
extern unsigned long overlay_plane_mask;
#define MAX_COLORS 2
#define MAX_PLANES 1
#define MAX_CELLS 4       /* MAX_COLORS * 2 ^ MAX_PLANES */
#define CANNOT_OVERLAY 0
#define CAN_OVERLAY 1
int
get_colors()
{
   int default_depth;
   static char *name[] = {"Red", "Yellow", "Green", "Green"};
   XColor exact_defs[MAX_CELLS];
   Colormap default_cmap;
   int ncolors = 4;
   int plane_masks[MAX_PLANES];
   int colors[MAX_COLORS];
   int i;
   XVisualInfo visual_info;
   int class;
   default_depth = DefaultDepth(display, screen_num);
   default_cmap   = DefaultColormap(display, screen_num);
   if (default_depth == 1) {
      /* Must be StaticGray, use black and white */
      background_pixel = WhitePixel(display, screen_num);
      foreground = BlackPixel(display, screen_num);
      printf("using black and white\n");
      return(CANNOT_OVERLAY);
   }
   if (!XMatchVisualInfo(display, screen_num, default_depth,
         PseudoColor, &visual_info)) {
      if (!XMatchVisualInfo(display, screen_num, default_depth,
            DirectColor, &visual_info)) {
         /* No PseudoColor or TrueColor visual available at
          * default_depth; some applications might try for a
          * GrayScale visual here if they can use gray to
          * advantage, before giving up and using black and white */
         background_pixel = WhitePixel(display, screen_num);
         foreground = BlackPixel(display, screen_num);
         printf("using black and white\n");
         return(CANNOT_OVERLAY);
      }
   }
   /* Got PseudoColor or TrueColor visual at default depth */
   /* The visual we found is not necessarily the default visual, and
    * therefore it is not necessarily the one we used to create our
    * window; however, we now know for sure that color is supported,
    * so the following code will work (or fail in a controlled way) */
      if (XAllocColorCells (display, default_cmap, False, plane_masks,
         1, colors, 2) == 0) {
      /* Can't get enough read/write cells to overlay;
       * try at least to get three colors */
         if (XAllocColorCells (display, default_cmap, False,
            plane_masks, 0, colors, 3) == 0) {
         /* Can't even get that; give up and use black and white */
                     background_pixel = WhitePixel(display,
                     screen_num);
                     foreground = BlackPixel(display, screen_num);
         printf("using black and white\n");
         return(CANNOT_OVERLAY);
      }
      else
         ncolors = 3;
   }

   /* Allocated three or four colorcells successfully, now set their
    * colors -- three and four are set to the same RGB values */
   for (i = 0; i < ncolors; i++)
   {
      if (!XParseColor (display, default_cmap, name[i],
            &exact_defs[i])) {
         fprintf(stderr, "basic: color name %s not in database",
               name[i]);
         exit(0);
      }
      /* This needed before calling XStoreColors */
      exact_defs[i].flags = DoRed | DoGreen | DoBlue;
   }
   printf("got RGB values\n");
   /* Set pixel value in struct to the allocated ones */
   exact_defs[0].pixel = colors[0];
   exact_defs[1].pixel = colors[1];
   exact_defs[2].pixel = colors[0] | plane_masks[0];
   exact_defs[3].pixel = colors[1] | plane_masks[0];
   /* This sets the color of the read/write cells */
   XStoreColors (display, default_cmap, exact_defs, ncolors);
   printf("stored colors\n");
   background_pixel = exact_defs[0].pixel;
   foreground = exact_defs[1].pixel;
   if (ncolors == 4) {
      overlay_pixel_1 = exact_defs[2].pixel;
      overlay_pixel_2 = exact_defs[3].pixel;
      overlay_plane_mask = plane_masks[0];
      printf("set can\n");
      return(CAN_OVERLAY);
   }
   else {
      /* This must be used as a normal color, not overlay */
      overlay_pixel_1 = exact_defs[2].pixel;
      printf("set can't\n");
      return(CANNOT_OVERLAY);
   }
}

Implementing this is just like overlays: one buffer is the overlay, and the other is what we have previously considered static graphics. The only change in double buffering is that you draw into the invisible buffer by using pixel values that do not change the visible colors in the visible buffer. Then you use XFillRectangle() to set or clear all pixels in the plane or planes that distinguish between the visible and invisible buffers. The price of doing this is that you have much fewer colors in each picture. For example, on an 8-plane system, you can only use 16 colors since you have two buffers of 4 bits each. (Also note that there is a double-buffering extension in progress in the X Consortium.)

Using XAllocColorPlanes()

Note that for applications like a paint mixing program, in which you have three bars for the three primary colors and a palette that shows the mixed color, you would not use XAllocColorPlanes(). The correct way to implement this is to allocate a single read/write color for the palette and to change it dynamically. (If the primary colors are displayed, they should be allocated using read-only colors.)

The piece of code shown in Example 7-8 is similar to Example 7-4 but it uses XAllocColorPlanes(). It is somewhat sketchy, because real applications that use XAllocColorPlanes() are complicated.

After allocating colors with XAllocColorPlanes(), you can then use XStoreColors() to set the colors. When nred, ngreen, and nblue are each 8, only one call to XAllocColorPlanes() and one call to XStoreColors() are necessary to allocate and set all 16 million colors of an entire 24-plane colormap.

Using XAllocColorPlanes() to allocate colorcells for DirectColor

#define PIXELS 256
Display *display;
int screen_num;
int contig = False;         /* Noncontiguous planes */
unsigned long pixels[PIXELS];   /* Return of pixel values */
/* Number of independent pixel values allocate */
unsigned int ncolors = PIXELS;
/* Need PIXELS * 2 ^ maxplanes defs, where maxplanes
 * is the largest of nred, ngreen, and nblue */
XColor defs[2048];
/* Number of planes to allocate for each primary */
unsigned int nreds = 3, ngreens = 3, nblues = 2;
/* Returned masks, which bits of pixel value for each primary */
unsigned long red_mask, green_mask, blue_mask;
Colormap colormap;
Status status;
/* Open display, etc. */
/* Get or create large DirectColor colormap */
while (status = XAllocColorPlanes(display, colormap,
      contig, pixels, ncolors, nreds, ngreens, nblues,
      &red_mask, &green_mask, &blue_mask) == 0) {
   /* Make contig False if it was True; reduce value of
    * ncolors; reduce value of nreds, ngreens, and/or
    * nblues; or try allocating new map; break when
    * you give up */
   break;
}
if (status == 0) {
   fprintf(stderr, "%s: couldn't allocate requested colorcells",
         argv[0]);
   exit(-1);
}
/* Define desired colors in defs */
while (status = XStoreColors(display, colormap, defs,
      ncolors) == 0) {
   fprintf(stderr, "%s: can't store colors", argv[0]);
   /* Try to fix problem here, exit or break */
   exit(-1);
}

/* Draw your shaded stuff! */

Getting Complete Visual Information

XGetVisualInfo() returns a list of visual structures that match the attributes specified by template and mask arguments. The template is an XVisualInfo structure with members set to the required values, and the mask indicates which members are matched with the list of available visuals. By passing an empty template structure, you can get a complete list of XVisualInfo structures.

The XVisualInfo Structure

The XVisualInfo structure is shown in Example 7-9.

The XVisualInfo structure

typedef struct {
   Visual *visual;
   VisualID visualid;
   int screen_num;
   unsigned int depth;
   int class;
   unsigned long red_mask;
   unsigned long green_mask;
   unsigned long blue_mask;
   int colormap_size;     /* Same as map_entries member of Visual */
   int bits_per_rgb;
} XVisualInfo;

The visualid member is not normally needed by applications.

As discussed earlier, the class member specifies whether the screen is to be considered color or monochrome and changeable or immutable. The class member can be one of the constants DirectColor, GrayScale, PseudoColor, StaticColor, StaticGray, or TrueColor.

The red_mask, green_mask, and blue_mask members are used only for the DirectColor and TrueColor visual classes, where there is a separate map for each primary color. They define which bits of the pixel value index into the colormap for each primary color. Each mask has one contiguous set of bits, with no bits in common with the other masks. These values are zero for monochrome and most four- to eight-plane color systems.

The colormap_size member of the structure tells you how many different pixel values are valid with this visual. For a monochrome screen, this value is two. For the default visual of an eight-plane color system, this value is typically 254 or 256 (two colors are often reserved for the cursor). For DirectColor and TrueColor, colormap_size will be the number of cells for the biggest individual pixel subfield. The colormap_size member is the same as the map_entries member of the visual structure.

The bits_per_rgb member specifies how many bits in each of the red, green, and blue values in a colorcell are used to drive the RGB gun in the screen. For a monochrome screen, this value is one. For the default visual of an eight-plane color system, this value is typically eight. The pixel subfields (the red, green, and blue values in each colorcell) are 16-bit unsigned short values, but only the highest bits_per_rgb bits are used to drive the RGB gun in the screen. This number corresponds the number of bits of resolution in the Digital to Analog Converter (DAC) in the screen hardware.

Example of Choosing a Visual

Code to match visuals

#include <X11/Xlib.h>
#include <X11/Xutil.h>
visual()
{
Display *display;
Colormap colormap;
Window window;
XSetWindowAttributes attributes;
unsigned long valuemask;
int screen_num;
   .
   .
   .
XVisualInfo vTemplate;    /* Template of the visual we want */
XVisualInfo *visualList;  /* List of XVisualInfo structs that
                           * match */
int visualsMatched;       /* Number of visuals that match */
   .
   .
   .
/* Set up the XVisualInfo template so that it returns a list
 * of all the visuals of depth 8 defined on the current screen
 * by the X server */
vTemplate.screen = screen_num;
vTemplate.depth = 8;
visualList = XGetVisualInfo (display, VisualScreenMask |
      VisualDepthMask, &vTemplate, &visualsMatched);
if ( visualsMatched == 0 )
   fatalError ("No matching visuals\n");
/* Create a colormap for a window using the first of the
 * visuals in the list of XVisualInfo structs returned by
 * XGetVisualInfo */
colormap = XCreateColormap (display, RootWindow(display, screen_num),
   visualList[0].visual, AllocNone);
/* Must specify colormap attribute if using nondefault visual */
attributes.colormap = colormap;
valuemask |= CWColormap;
   .
   .
   .
window = XCreateWindow (display, RootWindow(display, screen_num),
   x, y, width, height, border_width, vTemplate.depth,
   InputOutput, visualList[0].visual, valuemask, &attributes);
XSetWindowColormap(display, window, colormap);
/* All done with visual information; free it */
XFree(visualList);
   .
   .
   .
} /* End routine */

The GrayScale Visual

The color names "gray0" through "gray100", spelled with an "e" or an "a", can be used with XParseColor() to get RGB values for various grays.

You should set the red, green, and blue values to be equal. Some servers only use one of the values, and others combine all three according to the NTSC standard that makes color television signals work on black-and-white televisions:

intensity = (.30 * red) + (.59 * green) + (.11 * blue)

Standard Colormaps

X defines a set of properties that contain information describing commonly used colormaps. An application reads these properties by calling XGetRGBColormaps(). This call returns an XStandardColormap structure that contains enough information so that the application can calculate the colors in every colormap cell (or a certain range within the colormap). This structure may also include the ID of a colormap matching this description that was created by the window manager or another client. The X distribution from MIT includes a standard client, xstdcmap, that creates the standard colormaps. The user can arrange for xstdcmap to be invoked when the X server starts up, or in the user's .xsession or .xinitrc file. If this program has not been run, the application can create a new colormap and use the information in the standard colormap properties to allocate and set the colors according to the information in the property. There are functions in the Xmu (miscellaneous utilities) library for allocating standard colormaps.

But how does the sharing work? After creating this colormap, the application (or xstdcmap) sets the ID of the created colormap into the colormap field of the XStandardColormap structure and then calls XSetRGBColormaps(). This resets the property, so that the next time another client calls XGetRGBColormaps(), the colormap field of the returned structure will actually contain the ID of the appropriate colormap.

Therefore, although an application must have the code to create, allocate, and set colors in a standard colormap to be robust, in some cases this code will not be executed because some other client will have already done the work. After calling XGetRGBColormaps(), if the colormap field is zero, the application must create the colormap. Otherwise, the colormap field holds the ID of an appropriate colormap.

When an application uses standard colormaps, two (or more) instances of the application can run at the same time without increasing the load on the system caused by creating multiple copies of the same colormap. Applications that do not use standard colormaps will end up creating separate but identical colormaps. The window manager will switch these in and out of the hardware colormap whenever a different instance is in use. Although nothing on the screen will change color because both the colormaps are identical, the server will be performing unnecessary installing and uninstalling, and the extra colormaps will waste server memory.

In some cases, the window manager or even the server will create one or more standard colormaps. This does not change how applications work at all. Applications do not care whether it was the window manager, the server, or some other client that created a standard colormap.

If your application does not create or use a custom colormap, you can skip this section if pressed for time.

Applications can also use the knowledge about a standard colormap to optimize the process of figuring out which existing pixel values correspond to required colors and which colors must be allocated and set from scratch.

The Standard Colormap Properties

Properties were introduced in Section 2.1.4, "Properties and Atoms." For a more complete description of properties, see Section 12.1, "Properties and Atoms."

In the call to XGetRGBColormaps() you specify one of these atoms like XA_RGB_BEST_MAP (or, if necessary, one unique to your application).

The following list names the atoms and describes the colormap associated with each one:

XA_RGB_DEFAULT_MAP

This property defines part of the system default colormap. This colormap may be initially completely unallocated, or it may contain a selection of read-only colorcells with the RGB values from the color database and a few unallocated cells for use by applications that need read/write cells. A typical allocation of the XA_RGB_DEFAULT_MAP on eight-plane screens is all the colors produced from any combination of six reds, six greens, and six blues. This gives 216 uniformly distributed colors and leaves 40 for other programs or for special purpose colors for text, borders, and so on. A typical allocation for the XA_RGB_DEFAULT_MAP on 24-plane screens is 64 reds, 64 greens, and 64 blues. This gives about one million uniformly distributed colors (64 intensities of 4096 different hues) and leaves lots of colorcells available for other purposes.

XA_RGB_BEST_MAP

This property defines the "best" RGB colormap available on the screen. Of course, this is a subjective evaluation. Many image-processing and 3-D programs need to use all available colormap cells and to distribute as many perceptually distinct colors as possible over those cells. In this case, there may be more green values available than red and more green or red than blue.

On an eight-plane DirectColor visual, XA_RGB_BEST_MAP is usually a 3/3/2 allocation. On a 24-plane DirectColor visual, XA_RGB_BEST_MAP is usually an 8/8/8 allocation. On other screens, XA_RGB_BEST_MAP is purely up to the implementor of the server.

XA_RGB_RED_MAP, XA_RGB_GREEN_MAP, XA_RGB_BLUE_MAP

These properties define all-red, all-green, and all-blue colormaps, respectively. These maps are used by programs that make color-separated images. For example, a user might generate a full color image on an eight-plane screen by rendering an image once with high color resolution in red, once with green, and once with blue and exposing a single frame in a camera with three images.

XA_RGB_GRAY_MAP

This property describes the "best" gray-scale colormap available on the screen.

The XStandardColormap Structure

The XStandardColormap structure

typedef struct _XStandardColormap {
   Colormap colormap;
   unsigned long red_max, green_max, blue_max;
   unsigned long red_mult, green_mult, blue_mult;
   unsigned long base_pixel;
   VisualID visualid;         /* Added in R4:  ICCCM version 1 */
   XID killid;                /* Added in R4:  ICCCM version 1 */
} XStandardColormap;

• The colormap member is the ID of a colormap created by the XCreateColormap() function or the default colormap. This ID can be used to install a virtual colormap into the hardware colormap.
• The red_max, green_max, and blue_max fields give the maximum red, green, and blue values, respectively. A typical allocation that provides 63 = 216 read-only, shareable colors in a PseudoColor colormap on a standard eight-plane workstation is red_max = 5, green_max = 5, and blue_max = 5. This leaves 40 cells available for special colors and private, nonshareable purposes. On a 24-plane workstation, there would be eight bits available for each color in a TrueColor visual, which would allow 256 shades of each primary color. In this case, red_max = 255, green_max = 255, and blue_max = 255. This map would include 2563 = 16.38 million total colors.
• The red_mult, green_mult, and blue_mult fields scale each pixel subfield into the proper range in the 16-bit RGB value in the colorcell with the range 0 to 65535. The red pixel subfield is moved red_mult bits toward the most significant bit of the pixel value.For a 3/3/2 DirectColor allocation (eight reds, eight greens, four blues), red_mult might be 32, green_mult might be 4, and blue_mult might be 1 (as shown in Figure 7-5). These effectively move the red value into the most significant bits of the RGB value in the colorcell, the green into the middle, and the blue into the least significant bits. This arrangement is arbitrary but useful. For a six-colors-each allocation, which must be PseudoColor since the planes cannot be evenly allocated to separate primaries, red_mult might be 36, green_mult might be 6, and blue_mult might be 1.
• The base_pixel field gives the base value that is added to the pixel value calculated from the RGB values and scale factors. Usually the base_pixel is obtained from a call to the XAllocColorPlanes() function.
• The visualid field is the ID of a server resource associated with each visual, of type VisualID. You will need this ID only if you intend to use standard colormaps. This field was added in R4 because only with this information can standard colormaps be used with other than the default visual. Prior to R4, standard colormaps were not in wide use, partly because they could only be used with the default visual.
• The killid field returns a number that is used if the application needs to free the colormap for some reason. If killid is greater than one, then the resources should be freed by calling XKillClient() with the killid field as the argument. If killid is one, then the resources should be freed by calling XFreeColormap() with the colormap field as the argument. If killid is zero, then no attempt should be made to free the resources.

The 3/3/2 Standard Colormap

red    0-7    thus    red_max    =  7
green  0-7            green_max  =  7
blue   0-3            blue_max   =  3

Shifting pixel subfields into pixel value

 In a DirectColor system like this, the multiples are equal to 2n, where n is their lowest plane or bit position. If the red, green, and blue were stored in a different order, the multiples would not be 32, 4, 1 but would still be calculated from the above description and formula. The 3/3/2 standard colormap allocation is fairly standard.

Creating and Using a Standard Colormap

Example 7-12 gets information about the XA_RGB_BEST_MAP standard colormap, creates it if no other client already has, calculates pixel values from it, and sets the colormap window attribute of the window. This example gives up and falls back on read-only colorcell allocation if the standard colormap property is not defined by the server or if creating a colormap returns the default colormap (which happens on systems with an immutable hardware colormap).

Code to create and use XA_RGB_BEST_MAP

  .
  .
  .
#define USE_DEFAULT_COLORMAP 1
#define USE_STANDARD_COLORMAP 0
void main(argc, argv)
int argc;
char **argv;
{
    XStandardColormap *best_map_info;
    XColor *exact_defs;
    XSetWindowAttributes attrib;
    unsigned long attribmask;
    int i, j, k, l;
    int ncells;
    XVisualInfo *vlist, vinfo_template, *v;
    int num_vis;
    int count;
    Visual *visual;
    int strategy = USE_STANDARD_COLORMAP;
      .
      .
      .
    /* Open display */
    visual = DefaultVisual(display, screen_num);
    if (XGetRGBColormaps(display, RootWindow(display, screen_num),
            &best_map_info, &count, XA_RGB_BEST_MAP) == 0) {
        printf("%s: RGB_BEST_MAP colormap property not set.\n", argv[0]);
        /* Give up standard colormaps; use one of the
         * basic color strategies */
        get_colors();
        strategy = USE_DEFAULT_COLORMAP;
    }
    else if (best_map_info->colormap) {
        /* Someone else created the map we need; make sure
         * it's valid, then we'll use it below */
        if (best_map_info->red_max == 0) {
            printf("%s: RGB_BEST_MAP colormap property is set\n",
                    argv[0]);
            printf("but is missing data.\n");
            strategy = USE_DEFAULT_COLORMAP;
        }
        else {
            printf("stnd colormap ID: %d, best_map_info->colormap);
            attrib.colormap = best_map_info->colormap;
       }
    }
    else if (best_map_info->visualid == 0) {
        printf("%s: Standard colormap property is set\n", argv[0]);
        printf("but is missing data.");
        /* Some systems define the properties but don't
         * place any data in them; this is a server bug,
         * but we'll check for it anyway */
        /* Fall back on a basic color strategy */
        strategy = USE_DEFAULT_COLORMAP;
    }
    else {
        /* Got information, but the described colormap
         * has not been created yet; create it and
         * allocate all cells read/write */
        /* XCreateColormap requires a visual argument
         * (pointer to a Visual structure); however, the
         * XStandardColormap structure returns a VisualID,
         * which might not be the default visual;
         * Converting between these two is painful */
        vlist = XGetVisualInfo(display, VisualNoMask,
                &vinfo_template, &num_vis);
        for (v = vlist; v < vlist + num_vis; v++) {
            if (v->visualid == best_map_info->visualid) {
                visual = v->visual;
                break;
            }
        }
        best_map_info->colormap = XCreateColormap(display,
                RootWindow(display, screen_num), visual, AllocAll);
        if (best_map_info->colormap ==
                DefaultColormap(display, screen_num)) {
            printf("%s: hardware colormap is immutable:\n",
                    argv[0]);
            printf("cannot create new colormap.\n");
        }
        attrib.colormap = best_map_info->colormap;
        ncells = best_map_info->base_pixel +
                ((best_map_info->red_max + 1) *
                (best_map_info->green_max + 1) *
                (best_map_info->blue_max + 1));
        exact_defs = (XColor *) calloc(sizeof(XColor), ncells);
        /* Permute the levels of red, green, and blue */
        l = best_map_info->base_pixel;
        for (i = 0; i < best_map_info->blue_max; i++) {
             for (j = 0; j < best_map_info->blue_max; j++) {
                  for (k = 0; k < best_map_info->blue_max; k++) {
                       exact_defs[l].red = 0xFFFF * k /
                                 best_map_info->red_max;
                       exact_defs[l].green = 0xFFFF * j /
                                 best_map_info->green_max;
                       exact_defs[l].blue = 0xFFFF * i /
                                 best_map_info->blue_max;
                    l++;
                    }
             }
        }
        XStoreColors (display, best_map_info->colormap, exact_defs, ncells);

        /* If to be used in a window not created with the
         * default visual, must create the window first and
         * use instead of RootWindow in this call; here we
         * assume the default visual */
        XSetRGBColormaps(display, RootWindow(display, screen_num),
                &best_map_info, count, XA_RGB_BEST_MAP);
    }
    if (strategy == USE_STANDARD_COLORMAP) {
        /* We must not have called get_colors above,
         * must be using standard colormaps strategy */
        /* Note that we act like we have already allocated pixel
         * pixel values, even though actually another client did */
        background_pixel = best_map_info->base_pixel  +
            (best_map_info->red_max * best_map_info->red_mult) +
            (best_map_info->green_max * best_map_info->green_mult) +
            (best_map_info->blue_max * best_map_info->blue_mult);
        attribmask = CWBackPixel | CWColormap;
        foreground_pixel = (best_map_info->green_max *
            best_map_info->green_mult / 2) +
            best_map_info->base_pixel;

        border_pixel = (best_map_info->blue_max *
            best_map_info->blue_mult / 2) +
            best_map_info->base_pixel;
    }
    /* Create opaque window */
    win = XCreateWindow(display, RootWindow(display,screen_num), x,
            y, width, height, borderwidth, DefaultDepth(display, screen_num),
            InputOutput, visual, attribmask, &attrib);
      .
      .
      .
}

If the colormap field of best_map_info is nonzero, it should be the ID of a standard colormap that another client has created. Your application can immediately proceed to use the pixel values in this colormap as though your application had already allocated them read-only, even though in reality some other client allocated them read/write.

On the other hand, if the colormap field is zero, your application needs to create, allocate, and set the values of the standard colormap itself. You allocate the cells read/write, because this allows you to explicitly set the RGB values of each pixel value. Even though you allocate the cells read/write, you should use them as if they were read-only, so that other applications can share them after you reset the XA_RGB_BEST_MAP property to include the new colormap ID. As this suggests, a read/write cell, even though described earlier as being private and changeable by that one client, can be public if all the applications agree not to change its RGB values.

The algorithm used to store RGB values into the cells in the colormap is somewhat arbitrary. Conventions for it will probably be adopted by the X Consortium when there is more interest in standard colormaps. Any algorithm is good enough to allow two instances of the same application to share a colormap. But for two different applications to share the colormap, each must know exactly what RGB values the other would place in the colormap if the other were run before the colormap was created.

The XGetVisualInfo() call is described in Section 7.6, "Getting Complete Visual Information."

If you pass AllocAll to XCreateColormap(), you do not need to make an XAllocColorCells() call to allocate all the cells read/write. However, you can use AllocAll only if you intend the entire colormap to be read-only to all clients. Some clients want a few cells preserved to be rewriteable for dynamically changing colors. If yours is that way, you must use AllocNone and then call XAllocColorCells() once to create the standard portion of the colormap and again to allocate the cells your application will treat as private.

Once the colormap window attribute of a window is set, the window manager will take care of installing the colormap. When there is only one hardware colormap, the window manager usually installs an application's colormap when that application contains the pointer (for real-estate type window managers) or is given the keyboard focus (for click-to-type style window managers).

When a window manager creates a standard colormap, it can use a slightly different technique to make sure that the standard colormap remains defined even after the window manager exits. Assuming that it has already checked to see whether some other client has created a standard colormap and none has, it performs the following sequence of steps:

• Create a new connection to the same server.
• Determine the color capabilities of the screen. Choose a visual.
• Create a colormap (not required for XA_RGB_DEFAULT_MAP).
• Call XAllocColorPlanes() or XAllocColorCells() to allocate cells in the colormap (if did not use AllocAll flag when creating the colormap).
• Call XStoreColors() to store appropriate color values in the colormap.
• Fill in the descriptive fields in the XStandardColormap structure, including the ID of the created colormap.
• Call XSetRGBColormaps() to set the property on the root window. The killid field should be set to the colormap ID.
• Use XSetCloseDownMode() to make the resource permanent.
• Close the new connection to the server.

RGB-to-Pixel Conversion

Consider a 3-D display program that draws a smoothly shaded sphere. At each pixel in the image of the sphere, the program computes the intensity and color of light reflected to the viewer. The result of each computation is a triple of red, green, and blue coefficients in the range 0.0 to 1.0. To draw the sphere, the program needs a colormap that provides a large range of uniformly distributed colors. The colormap must be arranged so that the program can convert its RGB triples into pixel values very quickly, because drawing the entire sphere will require many such conversions. An example of one such calculation is shown in Example 7-13. Example 7-12 demonstrated how to do this for integral RGB values.

Calculating pixel values from floating point RGB values

XStandardColormap best_map_info;
float red, green, blue;
unsigned long pixelvalue;
pixelvalue = best_map_info.base_pixel +
   ((unsigned long)(0.5 + (red * best_map_info.red_max)) *
        best_map_info.red_mult) +
   ((unsigned long)(0.5 + (green * best_map_info.green_max)) *
        best_map_info.green_mult) +
   ((unsigned long)(0.5 + (blue * best_map_info.blue_max)) *
        best_map_info.blue_mult);

base_pixel <= pixel_value < (red_max * red_mult) + base_pixel

pixel_value = gray * red_mult + base_pixel;

   gray           =   the gray value you desire (0 to red_max)
   red_mult       =   value from XStandardColormap structure
   base_pixel     =   value from XStandardColormap structure