In vanilla Tcl/Tk, there is a single pool of commands, so the access command for a widget is the same as the window name. When a widget is created within a namespace, however, its access command is installed in that namespace, and should be accessed outside of the namespace using a qualified name. For example,
namespace foo {
namespace bar {
button .b -text "Testing"
}
}
foo::bar::.b configure -background red
pack .b
Note that the window name ".b" is still used in conjunction with commands like pack and destroy. However, the access command for the widget (i.e., name that appears as the first argument on a command line) must be more specific.
The "winfo command" command can be used to query the fully-qualified access command for any widget, so one can write:
[winfo command .b] configure -background red
and this is good practice when writing library procedures. Also, in conjunction with the bind command, the "%q" field can be used in place of "%W" as the access command:
bind Button <Key-Return> {%q flash; %q invoke}
While this behavior makes sense from the standpoint of encapsulation, it causes problems with existing Tcl/Tk applications. Many existing applications are written with bindings that use "%W". Many library procedures assume that the window name is the access command.
The itcl::purist variable controls a backward-compatibility mode. By default, this variable is "0", and the window name can be used as an access command in any context. Whenever the unknown procedure stumbles across a widget name, it simply uses "winfo command" to determine the appropriate command name. If this variable is set to "1", this backward-compatibility mode is disabled. This gives better encapsulation, but using the window name as the access command may lead to "invalid command" errors.