BookReview: The Programmer's Apprentice
by Charles Rich, Richard C. Waters, ACM Press, 1990, 0-201-52425-2

The theory: a programmer needs an apprentice to help with the mundane tasks. Programmers must do the higher level connections and the "checking up" on the apprentice while the apprentice deals with the details and low level checking up (e.g. do these constraints match the code?, are there contradictions?). The user is expected to be an expert software engineer (chief programmer). The apprentice will work with all levels from requirements to coding.

Knowledge Based Emacs (KBEmacs) has been implemented. It is an extended editor for editing programs with knowledge based commands. A plan calculus is developed for representing program cliches (program fragments). A cliche is how the engineer thinks. Engineers do not go to basic principles when programming, they use predefined cliches (roles, plans). The calculus is data flow oriented for language independence and ease of connections.

Much of the book is devoted to the calculus and KBEmacs. Some discussion at the end of the real state of the art and the problems. There are many plans and interesting insights to CASE. The problem is that a true apprentice doesn't exist and is hard to do (e.g. the plan calculus doesn't handle all programs).

I particularly liked the tricks with dataflow. This relates a little bit to my work with G and, in a sense, justifies the data flow approach.

Via Rob 1990