Learning GNU Emacs, Second Edition

A Note on Good X Programming Style

In the section describing frame objects earlier on, we noted that several frame-related Emacs LISP functions duplicate operations such as window iconification that are normally driven through your X Window System manager. As you study the Emacs LISP Reference Manual, you’ll find that there are other such functions we haven’t covered here--for example, it’s possible to read and write the X server selection buffers directly from Emacs LISP. Although functions like these are made available in Emacs LISP in accordance with its design philosophy, you should think twice--and then think again—before using them in an Emacs LISP program that isn’t itself driven by an X input event loop. Conventional interactive programs (including Emacs on character terminals) have a modal I/O style: they’re designed to be used one at a time, and they explicitly solicit input from the user when they’re good and ready for it. A lot of the program’s state is determined by where the user is in the overall execution sequence. The interaction profile of such programs looks like a collection of little I/O loops in order, each one expecting a subset of input events potentially different from the previous.

In the X world, by contrast, almost all programs (and all good programs) are driven by one big nonmodal event loop that accepts a large range of X input events besides ordinary keystrokes, including mouse clicks, exposure events, and resizes. Very little of the program state ...