A photo is an image whose pixels can display any color or be transparent. A photo image is stored internally in full color (32 bits per pixel), and is displayed using dithering if necessary. Image data for a photo image can be obtained from a file or a string, or it can be supplied from C code through a procedural interface. At present, only GIF and PPM/PGM formats are supported, but an interface exists to allow additional image file formats to be added easily. A photo image is transparent in regions where no image data has been supplied.
Like all images, photos are created using the image create command. Photos support the following options:
When a photo image is created, Tk also creates a new command whose name is the same as the image. This command may be used to invoke various operations on the image. It has the following general form:
imageName option ?arg arg ...?Option and the args determine the exact behavior of the command.
Those options that write data to the image generally expand the size of the image, if necessary, to accommodate the data written to the image, unless the user has specified non-zero values for the -width and/or -height configuration options, in which case the width and/or height, respectively, of the image will not be changed.
The following commands are possible for photo images:
The photo image code is structured to allow handlers for additional image file formats to be added easily. The photo image code maintains a list of these handlers. Handlers are added to the list by registering them with a call to Tk_CreatePhotoImageFormat. The standard Tk distribution comes with handlers for PPM/PGM and GIF formats, which are automatically registered on initialization.
When reading an image file or processing string data specified with the -data configuration option, the photo image code invokes each handler in turn until one is found that claims to be able to read the data in the file or string. Usually this will find the correct handler, but if it doesn't, the user may give a format name with the -format option to specify which handler to use. In fact the photo image code will try those handlers whose names begin with the string specified for the -format option (the comparison is case-insensitive). For example, if the user specifies -format gif, then a handler named GIF87 or GIF89 may be invoked, but a handler named JPEG may not (assuming that such handlers had been registered).
When writing image data to a file, the processing of the -format option is slightly different: the string value given for the -format option must begin with the complete name of the requested handler, and may contain additional information following that, which the handler can use, for example, to specify which variant to use of the formats supported by the handler.
When a photo image is displayed in a window, the photo image code allocates colors to use to display the image and dithers the image, if necessary, to display a reasonable approximation to the image using the colors that are available. The colors are allocated as a color cube, that is, the number of colors allocated is the product of the number of shades of red, green and blue.
Normally, the number of colors allocated is chosen based on the depth of the window. For example, in an 8-bit PseudoColor window, the photo image code will attempt to allocate seven shades of red, seven shades of green and four shades of blue, for a total of 198 colors. In a 1-bit StaticGray (monochrome) window, it will allocate two colors, black and white. In a 24-bit DirectColor or TrueColor window, it will allocate 256 shades each of red, green and blue. Fortunately, because of the way that pixel values can be combined in DirectColor and TrueColor windows, this only requires 256 colors to be allocated. If not all of the colors can be allocated, the photo image code reduces the number of shades of each primary color and tries again.
The user can exercise some control over the number of colors that a photo image uses with the -palette configuration option. If this option is used, it specifies the maximum number of shades of each primary color to try to allocate. It can also be used to force the image to be displayed in shades of gray, even on a color display, by giving a single number rather than three numbers separated by slashes.
The photo image type was designed and implemented by Paul Mackerras, based on his earlier photo widget and some suggestions from John Ousterhout.