Subscribe
Subscribe
MY AMERICAN SCIENTIST
LOG IN! REGISTER!
SEARCH
 
Logo IMG
HOME > PAST ISSUE > May-June 2003 > Article Detail

COMPUTING SCIENCE

The Post-OOP Paradigm

Brian Hayes

Automating Automation

Surely the most obvious place to look for help with programming a computer is the computer itself. If Fortran can be compiled into machine code, then why not transform some higher-level description or specification directly into a ready-to-run program? This is an old dream. It lives on under names such as generative programming, metaprogramming and intentional programming.

In general, fully automatic programming remains beyond our reach, but there is one area where the idea has solid theoretical underpinnings as well as a record of practical success: in the building of compilers. Instead of hand-crafting a compiler for a specific programming language, the common practice is to write a grammar for the language and then generate the compiler with a program called a compiler compiler. (The best-known of these programs is Yacc, which stands for "yet another compiler compiler.")

Generative programming would adapt this model to other domains. For example, a program generator for the kind of software that controls printers and other peripheral devices would accept a grammar-like description of the device and produce an appropriately specialized program. Another kind of generator might assemble "protocol stacks" for computer networking.

Krzysztof Czarnecki and Ulrich W. Eisenecker compare a generative-programming system to a factory for manufacturing automobiles. Building the factory is more work than building a single car by hand, but the factory can produce thousands of cars. Moreover, if the factory is designed well, it can turn out many different models just by changing the specifications. Likewise generative programming would create families of programs tailored to diverse circumstances but all assembled from similar components.




comments powered by Disqus
 

EMAIL TO A FRIEND :

Of Possible Interest

Letters to the Editors: The Truth about Models

Spotlight: Briefings

Computing Science: Belles lettres Meets Big Data

Subscribe to American Scientist