if (...) {
if (...) {
...
} else {
if (...) {
...
} else if (...) {
...
} else {
...
}
}
and prefer this:
if (...)
{
if (...)
{
...
}
else
{
if (...)
{
...
}
else if (...)
{
..
}
else
{
...
}
}
}
The first example is confusing and hard to read. The second is clean and easy to read.
As to the question of optimization, that's a different matter. Lots of tricks can be used to compress code into fewer statements on a case by case basis:
if(pushbutton == 0)
{
led = 12;
}
else
{
led = 24;
}
I find the first version easy to read, and the second confusing and hard to read. I believe this is a matter of what a person is more accustomed to, and I learned on the first form and lately have been using the first form more. For me the extra brace below the keywords makes it hard to find the phrase of interest, and the expanded vertical space also makes it more difficult to line up indentation levels. But these visual cues are flexible, and I could learn similar scanning tricks for the second form.
I have worked with people that were dogmatic that one form was objectively better than another form, rather than it being a matter of local optimization in the brain; I have difficult working with these people because they tend to show similar rigidity and confusion of the subjective with the objective in other areas, and such small mindedness limits what type of engineering solutions one is able to come up with.
An early history of writing code on 24 x 80 terminals has trained me to prefer K&R style (where the bracket is on the same line as the if) over the expanded style. The goal was to have as much of the code on the screen as possible so vertical space was precious. Of course when you've got a 1080p monitor on its side you get all the vertical space you need and the open style does give you more room.
I strongly avoid however converting logical semantics into arithmetic ones. The LED example where:
led = (pushbutton * 12) + 12;
Makes me cringe.
You know that the person who wrote that assumed 'non-zero' meant it was pressed, not '1' means its pressed at some point the pushbutton is connected to bit 3 of the I/O port and the the kid was doing a byte read 'cuz it was easy' and now the value of pushbutton is 8 (which still would work on their code but not the refactored code) and led gets the value 108 and the code that takes the LED value turns on some random GPIO bit somewhere else because it can "only" get either 12 or 24 when now it can get a lot of different values. And a long escalation in manufacturing that the 'new software screws up the gizmo' turns out to be a side effect from refactoring. Been there, done that, wrote the after action report. No thanks!
First I understood that you were not seriously proposing this sort of I/O code, but as someone who has been around the block my share of times I know that such code both exists and is in production, almost exactly as written. The memories are like a war veteran who can't forget the trenches, I wince now and then :-).
Second, my comment was that code which has a 'boolean' flow is probably boolean code, and not arithmetic code. So any refactoring that 'assumes' boolean and uses arithmetic becomes a landmine for some future coder to trip over. MUCH better to refactor the code as such into:
led = (button)? 12 : 24;
Than "assuming" that true is '1' and false is '0'. When reading old code I've trained my self to read the statement
if (variable == 0) {
// this code assumes the variable is zero
} else {
// this code has to work for all non-zero values
}
Which may not have been the programmer's intent, but it is what they wrote.
I understand. This was about coding style and not about solving a particular problem correctly. Next time I might choose a better example. Maybe I'll include complete keyapad matrix scanning, debouncing and encoding code...just to be safe. (just kidding).
It's been a long, long time since I've checked button activity that way. These days andy time I do embedded work I tend to use either my own RTOS or a third party RTOS. In most of these cases UI events get turned in to messages in a message loop (sort of like programming for Windows). So it is likely that there's a big switch statement somewhere picking picking off messages off the FIFO and processing them accordingly. The low-level stuff happened somewhere else. The message would indicate the event that took place (UP, DOWN, CLICK, DOUBLE_CLICK) and parameters in the message packet would identify the button or buttons (in the case chords are enabled) that generated the event.
Fun stuff. Wrote this code years ago and keep reusing it. It simply works. Now doing a lot more iOS and web development. iOS can feel a lot like embedded development at times.
Good procedural code is factored into short functions, which would make some of the types of programs the article is alluding to easier to read.
Also, object oriented programming can be a good option for reducing cyclomatic complexity in a lot of different situations where the system you are modeling lends itself to a class structure. And CoffeeScript makes this easier also.
In C at least, pushbutton could be strictly greater than one and still be treated as true, but that would be a different result in the arithmetic versions of the statement. It could be made identical by using "!!pushbutton".
I agree, the second style is infinitely more readable. It clearly shows the hierarchy. Unfortunately it seems to be rare compared to the second one, as most people seem to want to hide the opening braces or something. Maybe because they come from ugly languages without braces, like Assembly, Python, Fortran, or Ruby...
I've programmed in everything from entering hex code into a hexadecimal keypad to assembly, Forth, APL and beyond. Sure, languages can influence style. I wouldn't go as far as calling any language "ugly". There's that initial phase of exposure when one might not be comfortable with the language constructs. And yes, at that moment in time it is natural to feel that a language might be "ugly". I certainly felt that way when learning Objective-C. Once you get on the other side and you start using the language for real work things can change. I don't hate any particular language but do enjoy some more than others. Forth, APL and Lisp were a lot of fun to work with.
I hate this style (from the article):
and prefer this: The first example is confusing and hard to read. The second is clean and easy to read.As to the question of optimization, that's a different matter. Lots of tricks can be used to compress code into fewer statements on a case by case basis:
can become or or etc.