Long code blocks inside if statements or for loops

问题:

This is a cross language question on coding style.

I have to work with a lot of code that has very long code blocks, sometimes hundreds of lines, inside if statements or for loops. The code is procedural.

Code like the following

if(condition){
    //hundreds of lines of code
}else if{
    //hundreds of lines of code
} else {
    //hundreds of lines of code
}

I have trouble navigating this code if I haven't seen it in a while because I constantly have to scroll back and forth to check which branch of the statement I'm in or whether I'm in a loop, or what the loop iterator is called.

My hunch is to put pieces of the long lines of code inside functions and call them from within the branches of the statements or inside the loops, so the loops and if trees are shorter and therefore more readable. I'd create functions that are sensible code-silos ofcourse, not just cut up the current code haphazardly.

But, and here's my question: is that a good idea? Or is it not a bad coding style to have hundreds of lines of code inside an if statement or a for loop? I'm not a really experienced programmer but I do appreciate clean code :)

Thanks!

Added: The hundreds of lines of code are not repetitive, most of the time. I understand and try to adhere to the DRY principle of course.

回答1:

It's generally a good idea to be able to see the whole of a method on one screen. If you have to scroll up and down too much (some would argue at all) then you run the risk of missing information.

So in general what you're proposing is a good idea. Though you needn't go as far as putting all the code in one branch into one method. There might be repeated sections of code so you could break it up as follows:

if(condition){
    CommonMethod();
    SpecificMethodA();
}else if{
    CommonMethod();
    SpecificMethodB();
} else {
    CommonMethod();
    SpecificMethodC();
}

as an example. Though the exact nature of the refactoring will depend totally on your code.

回答2:

From what you say, I'd refactor the contents of those blocks into separate procedures (giving them useful names and passing arguments as required).

if(condition){
    DoA();
}else if{
    DoB();
} else {
    DoC();
}

procedure DoA() {
    //hundreds of lines of code
}

procedure DoB() {
    //hundreds of lines of code
}

procedure DoC() {
    //hundreds of lines of code
}

Of course, it may be useful to do this inside of the DoX-procedures as well.

回答3:

It is awful coding style to have hundreds of lines in a single function. Have small functions, hopefully that fit in a single screen and that serve a single purpose, so that your code ends looking up like

if (condition) {
    func1(); //One purpose
    func2(); //Another purpose
    func3();
} else if (another condition) {
    func4();
} else {
    func5();
    func6();
    func3(); //wow I can reuse some code!
}

Don't go on the easy route to put the whole hundreds of lines of each case in a single function either. Try to isolate purposes (this function create a temporary file, for example) and assign functions to them. You will probably be surprised to see your total line count to go down, as there is very likely repeated functionality in those hundreds of lines.

回答4:

Have this in mind: You - or anyone else reading the code - should understand each function easily when reading it. If the function contains hundreds of lines of code it will never be easily understandable. If you find yourself making notes when reading to understand what the code does it needs refactoring.

The example you're showing is definitely of the kind that should be refactored. Each block of the if-clause should ideally call one function only.

if (condition) 
{ 
    doSomething(); 
} 
else if (anotherCondition) 
{   
    doThisOtherThing();
}
else
{   
    doThisThirdThing();
}

You see how easily understandable the code is? There is really nothing to understand there - except parhaps the condition, but if this is complex it should be factored out as a separate function too.

Now, the doSomething(), doThisOtherThing() and doThisThirdThing() function should not be the plain chunks of hundres lines of code. No function should ever be that long. It will always be difficult to understand that long functions. Try finding logical parts of your functions that can be factored out as separate functions and do so. If there are common parts that should be done in all of the cases a common function should be created and called from all of the doX-functions.

Note: You don't need code reuse to create a function. You can factor out something in a new function just to make the code more readable.

Good luck refactoring! :-)

回答5:

In addtion to the other answers: If you want to know more about this, try reading "Clean Code" by Robert C. Martin.

回答6:

In my opinion as a long time coder and code reviewer, the answer (as usual) is sometimes. Here are reasons I would split the code into separate functions:

1. The code would be more understandable if assigned function names. I find that comments are still rare, and creating a function name to describe a block of code is one way to document what the block of code does.
2. The parent function (as you note) would be more manageable if it weren't so long.
3. If the function can be designed intelligently, it might be useful elsewhere, or even help you re-factor the existing code into a smaller amount of code.

But there are also reasons to keep the code in the initial function:

1. It might be more of a pain to navigate back and forth to the child functions and back if the internal code is highly interactive with the parent function code.
2. The code could be less readable and maintainable if the internal code is highly interactive with the parent function code, having to pass many parameters to maintain the state of what the child function needs to process.
3. The code might not be unreadable or unmanageable in the first place if the internal code is relatively repetitive and simple. But keep in mind that if it's repetitive and simple, that might also make it a good candidate for re-factoring for other reasons.

回答7:

It's not a good coding style.
Breaking the code into smaller procedures makes it much more readable and maintanable.

Having said that. Don't immediately go and start cutting these up into smaller procedures.
Only change what you are reading regularily, or which you need to update.

Remember, every time you change code you need to test your change, so don't change needlessly, as every single time you change code, there is a potential to introduce a new bug.

回答8:

I once worked in a project where there was an official limit of 20 lines of code per function. It was hell!

I might be working with masks, moving data in or out of GUI elements, and then I'd be forced to write

transferFirst10Elements()
transferElements11Thru20()
transferElements21Thru30()
transferElements31Thru40()

and railed against our policy as an example of dogma taken too far.

Just to clarify, I think that initial function should indeed be broken up, but I don't agree with some other answerers that the breakdown should continue until there are no more fragments bigger than a screenful. There comes a point where managing zillions of tiny functions is more painful than managing somewhat fewer larger ones.

回答9:

The fact that you are experiencing problems whenever you revisit the code is clear enough indication that there is a problem.

Contrary to most of the the other suggestions here, I'd take a slightly different approach.

Trying to follow some of the suggested rules about where and how much to split may work, but it can also lead to situations where the breakdown itself is unmanageable. Instead of scrolling up and down to familiarise yourself again, you'd just be navigating to a bunch of smaller routines scattered throughout your source files.

Instead...

I suggest you try to focus on splitting your code into a number of methods where each method performs one task which is clearly indicated by the name of the method. Code blocks like these are usually large because they are doing multiple things. By splitting them in this way, you'll find that your code will naturally evolve into a more manageable style with less risk of the problem mentioned before.

It may also help to work in the other direction. Take one of your larger code blocks, try to (while ignoring the actual code) define what it does in a number of simpler steps, then code it accordingly. Remember, it must do one thing; but may require multiple steps to get the job done. This will provide natural breakdown of the large code-block.

NOTE: I would never advocate imposing line limits on routines that perform a single task; for example a routine to 'Populate Edit Controls' may be many lines because of the number of the number of controls, but the mistake some people make is to do 'simple calculations' in addition to populating the controls.

An additional benefit to this approach is that you're more likely to have routines that can be reused because they don't have 'excess baggage'.

回答10:

Procedures/Function are used to replace repetative code, where you only require minor details/values to change. This reduces the redundancy, and maintinance of the code. But yes you can also use it (with meaningfull names) to reduce such big code blocks.

回答11:

You should really consider decomposition of such a blocks into classes. OOP provides you with all techniques required. Template method pattern for example solves a problem with presence of CommonAction among condition-specific actions. This is just the tip of OOP iceberg.

回答12:

I'd second the template/pattern suggestion by archimed (sorry I'd vote you up, but my rep isn't that high yet apparently). I was thinking of a Strategy pattern myself, but Template would work out just as well.

I'm actually really surprised that it got voted down. Having done it both ways (via moving the code into methods versus implementing a pattern) and I've found that the amount of work to implement a pattern for those chunks of code is really not that much different. And as ya23 said, the long term benefits of having the pattern implemented are pretty substantial.

回答13:

Your idea of extracting the long code fragments inside of the "if" blocks is on the right track; however, I would start much smaller than that. Look for smaller bits of repeated code - I'd say 3-10 lines - that do something very understandable, and extract that to a method (which should be very easy to name). If possible, do the extracted using automated refactoring tools (to avoid errors) and add unit tests for the extracted method. Now, repeat this a couple of times, and you should see patterns and structures begin to emerge. You'll likely end up with a structure similar to that of a previous answer:

if (condition) {
    func1(); 
    func2(); 
    func3();
} else if (another condition) {
    func4();
} else {
    func5();
    func6();
    func3();
}