Tuesday, July 14, 2009

The Value Of Pictures In Software Design

There are some very good reasons why software engineers use visual communication to quickly and effectively transfer knowledge from one person to another.

While people have many different learning styles, and while everyone employs all of the styles to a greater or lesser degree, most people, or at least, enough people to matter, are predominantly visual learners. Various sources claim that around 60% of us are visual learners. Therefore, it's worthwhile to use visual techniques for this reason.

Visual communication transfers information at a very high rate compared with aural and textual communication. You can tell with a glance a system's structure, or lack thereof. A verbal description takes longer.

Visual communication helps the sender, too. That is, the person creating the graphical representation has to understand the system well enough to draw it. This applies to verbal and written communication as well, so visual does not necessarily hold an advantage over other forms, but it's certainly a valid approach.

Visual communication helps a newcomer to the team come up to speed and become productive more rapidly.

Furthermore, graphical representations of software systems can reveal flaws, voids, and redundancies that are not immediately obvious in verbal or written communication. How many times have you drawn a system diagram, only to see that there's something that can be cleaned up? If you have not done this, try it - it's a worthwhile exercise.

To illustrate the value of pictures, let me point you at one that someone else drew, one that helped me to quickly understand a software framework's design and intended usage. I'm talking about the design of Mina, Apache's ongoing effort to wrap the basic Java NIO components. I think the graphics they provide are a great example of what we should be doing on our own projects.

Why is it important to understand the need for visual communication in software development? I've noticed something unsettling on the last couple projects I've been on: no graphical representations of the systems. In each case, there was no perceived need or gain to having images. For those of us who have been on non-trivial projects and witnessed the indispensable benefits of this form of communication, this is a red flag.

This red flag is a reliable indicator of a project in trouble. The particular dysfunctions might take one of several forms, but most likely is a combination of them. I'll try to name a few here, without trying to make an exhaustive list.

First, it means that the system probably has no clear direction. The team doesn't know where it's headed, or at least doesn't have a common vision of the goal.A shiny new a idea comes along, it's legitimately cool, and we go down that track. And that's great - provided it complements the existing framework. We can't follow all of those cool ideas. Some we'll have to back-burner for another day or another project. A solid understanding of a system as a whole, bolstered by few good drawings, can help us stay disciplined and on the road toward our goal. The images help to remind us of the goal, and to keep us from working at cross-purposes.

A lack of graphical communication might indicate that the team is unable to create them. The system has grown disorganized and chaotic over time (that is separation of concerns is largely gone), and the team, however good they are, cannot walk up to a whiteboard or show on paper a cohesive, overall design. The discipline of striving to achieve the goal of always having something drawn (simple, not 200 pages), drives us to keep the system cohesive and well-organized.

A lack of drawing might indicate something far worse - that the team refuses to draw them. This might be from fear that the drawings will become stale (a legitimate, but addressable concern), or because it takes time from coding, or from a misapplied development philosophy, from plain and simple laziness, or because of a lack of experience building complex systems. Even with current development paradigms that eschew grotesquely large architectural documents, some documentation is essential.

It's this last statement that seems be key element on the last couple projects I've been on. Agile development philosophies encourage us to limit the amount of useless documentation that gets created. This is a worthy and noble goal. Sadly, some have twisted the intent of these goals, eliminating strong, time-tested tools from their arsenal, to the detriment of the projects and teams they represent.

No comments: