Software Design: The easy way is complicated

Wait... WHAT?

A programmer may find it easier to copy and paste a statement 10 times than to use a for loop if they have not done a for loop at least several times in the given language. The fact remains that tedious and repetitive work is easier than work that constantly differs by a large degree.

The real life story

A colleague of mine once made me change back something I had added, because she wanted to "keep things simple". What is it she wanted to change you ask? I had created a simple class with constant strings to name session variables(.Net). This way you could easily find troubles with session variables being misset or accidentally reused for something else, because instead of strings they are simple. One problem that arose is that the debugger refused to display the value of the session variable or use it in any expression, and thus it was removed, but at a cost.

What cost?

The costs are debugging and testing time. Keeping your code centralized is an important aspect of good design. If this was untrue there would be a sink in one room, and a toilet in the other. This would mean you can take a shit, but you have to walk accross the hall to wash your hands. Oversimplification of a problem requires narrowing of the problem domain, which means that you will be solving more oversimplified problems later. It can actually get to a point where renaming a variable can take days, and break entire system which rely on it.

Another example

If you don't use a for loop in your code, and copy and paste the same code snippet 30 times, you will most likely not have out of range indexes, and if your careful it will work perfectly. Then you leave the project and the next programmer comes along. They run into your hard coded loop and wonder: "Why did he do it this way?". They(and by they I mean me too, this is a real example) then have to sift through every single iteration making absolutely sure that there are no functional differences in any iteration. This almost bit me. A database field in a legacy system is limited to 10 characters. When the field names increment past 10, the last letter of the name is replaced by the 1 in the number. So you would have flag9, fla11 suffixing the field name. Another catch was that 10 was 0, so you would have a sequence: FLAG9, FLAG0, FLA11...

The problem was perfectly solvable with one for loop with 2 nested if statements. The programmer before me took the lazy route and made my route hell. I still worry a little if some business logic wasn't lost there, but I cry no tears for bad code. It needs to go. As soon as possible.

Complex implementation makes simple interface

To return to the copy and paste example, it is easy to copy and paste some little action everywhere you need it. A typical example is:

- Connect to Database
- Execute query
- Return result
- Close connection

Pasting snippets that do this is luckily not something I have seen often(except maybe in sloppy PHP). Writing a function that does the above will mean you never have to paste that snippet, and you can make global changes enhancing the way you data access layer functions.

Wrap up

I have seen many programmers wrestle with the same problem over and over again. I mean... I have seen the same programmer wrestle with the same problem he had a few days ago, because he solved the problem once, but did not wrap the answer up into a simple to remember way. In order to advance yourself and the systems you work on, you have to make sure you do the least work with maximum area effect.

Be rike the atom bomb, smaw bom, but spread rarge area.

Comments