Back when I was studying stuctured program design (and promptly forgot most of it and had to relearn it), I remember my instructor explaing that all programs had four steps:

1. Initialization
2. Input
3. Calculation
4. Output

Admittedly, that's a bit simplistic and I suspect that he was only familiar with procedural programming (as opposed to functional, OO, and logical). While I don't necessarily design my programs into four sections, they often break down that way.

What I usually do is start with the output. What do I want my end result to be? From that, I can determine what data I need to capture. Then, I go to the input section and figure out my data sources. With this, I have my beginning and end data and I find it relatively easy to work out the data flow (well, usually).

The data flow, to me, is the 'calculation' part. Frankly, I don't even worry about my data structures until I have some idea of what those calculations are. With that, I can often identify common elements to factor out. Once I know those, I design data structures to fit them.

I can't say that I follow this process religiously (ooh, I'm a bad monk), but what I prefer to do is something similar to this node. Also, Warnier-Orr design is something that I've enjoyed, though it often breaks down with object-oriented programming.

Cheers,
Ovid

Vote for paco!

Join the Perlmonks Setiathome Group or just click on the the link and check out our stats.