Friday, May 16, 2008

Patterns of Software

Patterns of Software (PDF) is an essay collection written by Dick Gabriel. The collection is, like software patterns, based on the work of Christopher Alexander. It is more a thoughtful analysis of Alexanders work, with comments related to software, than a practical implementation of his ideas as in the GoF-book. Please note that this is not my analysis of Alexander, whom I haven't read yet, but simply notes of Gabriels analysis of Alexander. Gabriel identifies three important concepts in software systems:

Compression - the ability to describe something succinctly by using the surrounding context. Examples are subclassing and closures. The benefit of compression is the elegance of avoiding unnecessary words that distract from the main issue. The downside is that the reader must know the surrounding context.

Habitability - the characteristic of source code that enables programmers to understand and change the it comfortably and confidently. It is related to Alexanders organic order which is defined as... "the kind of order that is achieved when there is a perfect balance between the needs of the whole and the need of the parts". If the programmers lose the sense of responsibility for the environment they live in hoe can they feel any sense of purpose there. Habitability enables piecemeal growth, small incremental modifications, a reality in software. An interesting and controversial idea on habitability is: "Habitability is not clarity. The danger of clarity is it's uncompromised beauty; it's really tough to improve uncompromised beauty. Clarity is dangerous."

Abstraction - is the thought process wherein ideas are distanced from objects. In programming the implementor of an abstraction can ignore the exact uses and the user of the abstraction can forget the details of the abstraction. Abstraction especially when it is take too far may limit the habitability of the software. It is important to keep the abstractions clear and intuitive. Abstracting is difficult and should be handled with care. Abstractions are part of the context that makes compression possible but at the cost of knowing the abstractions. Sometimes, when no good abstraction can be found, it is better to just write the common code.


Qwan- At the heart of Alexanders pattern language was his quest for the Quality without a name, Qwan. It is interesting to note that this central idea is not mentioned at all in the GoF-book. It is as if they found it to vague to be of any interest.

Gabriel does not claim to know what qwan is but mentions things about software that possesses it. Here are a few:

  • Its abstractions and modules are not too big. They make sense for themselves; they are whole.
  • If I look at any small part I can see what is going on, I don't need to look at other parts to understand it.
  • Everything about it seems familiar.
  • I can image changing it without fear of breaking it.
  • It is like a fractal in which every level of detail is locally coherent.
Patterns and Carpets - Alexander patterns contains people and their activities, this is because the buildings are meant to be lived in. This is also true for software; it is meant to be maintained. Another aspect of Alexander patterns was that their geometrical nature was fundamental.

Alexander also studied Turkish carpets since he found a beauty in them that he could not explain. He even published a book about them. In it he says:

  • The beauty of a structure comes from the fine detail at an almost microscopic level.
  • The beauty of the color of the carpet was due to the fact that the master dyer served a 15-year apprenticeship after which he was required to produce a color no one had seen before.
  • Fanatical precision is not required. It is only important to know what is important and what is not and to pay attention to what is important.
  • To get wholeness, you must leave things that don't matter rough and unimportant and give things that really matter deep attention.
  • Every successful center is made of a center surrounded by a boundary which is itself made by centers.
The solution to a large problem is a nested set of patterns.

Failure - Another interesting thing not mentioned in the software pattern literature is the fact that Alexander failed (his own words) when he tried to implement his ideas. The building that were built did not have the quality that he had expected.

No comments: