Simple Getter/Setter comments

Question

What convention do you use to comment getters and setters? This is something I've wondered for quite some time, for instance:

/**
 * (1a) what do you put here?
 * @param salary (1b) what do you put here?
 */
public void setSalary(float salary);

/*
 * (2a) what do you put here?
 * @return (2b)
 */
public float getSalary();

I always find I'm pretty much writing the exact same thing for 1a/b and 2a/b, something like 1a) Sets the salary of the employee, 1b) the salary of the employee. It just seems so redundant. Now I could see for something more complex you might write more in the (a) parts, to give context, but for a majority of the getters/setters out there the wording is almost exactly the same.

I'm just curious if, for the simple getters/setters its ok to only fill in either the (a) part OR the (b) part.

What do you think?

Solution 1

I usually just fill the param part for setters, and the @return part for getters:

/**
 * 
 * @param salary salary to set (in cents)
 */
public void setSalary(float salary);

/**
 * @return current salary (in cents, may be imaginary for weird employees)
 */
public float getSalary();

That way javadoc checking tools (such as Eclipse's warnings) will come out clean, and there's no duplication.

Solution 2

Absolutely pointless - you're better off without this kind of crap cluttering your code:

/**
 * Sets the foo.
 * 
 * @param foo the foo to set
 */
public void setFoo(float foo);

Very useful, if warranted:

/**
 * Foo is the adjustment factor used in the Bar-calculation. It has a default
 * value depending on the Baz type, but can be adjusted on a per-case base.
 * 
 * @param foo must be greater than 0 and not greater than MAX_FOO.
 */
public void setFoo(float foo);

Especially the explanation of what the property actually means can be crucial in domain models. Whenever I see a bean full of properties with obscure names that only investment bankers, biochemists or quantum physicists understand, and the comments explain that the setGobbledygook() method "sets the gobbledygook.", I want to strangle someone.

Solution 3

Generally nothing, if I can help it. Getters and setters ought to be self-explanatory.

I know that sounds like a non-answer, but I try to use my time for commenting things that need explanation.

Solution 4

I'd say only worry about commenting getters and setters if they have some sort of side effect, or require some sort of precondition outside of initialization (i.e.: getting will remove an item from a data structure, or in order to set something you need to have x and y in place first). Otherwise the comments here are pretty redundant.

Edit: In addition, if you do find a lot of side effects are involved in your getter/setter, you might want to change the getter/setter to have a different method name (ie: push and pop for a stack) [Thanks for the comments below]

Solution 5

Ask yourself what do you want people to see when the comments are viewed as JavaDocs (from a browser). Many people say that documentation is not necessary since it's obvious. This won't hold if the field is private (unless you explicitly turn on JavaDocs for private fields).

In your case:

public void setSalary(float s)
public float getSalary()

It's not clear what salary is expressed in. It is cents, dollars, pounds, RMB?

When documenting setters/getters, I like to separate the what from the encoding. Example:

/**
 * Returns the height.
 * @return height in meters
 */
public double getHeight()

The first line says it returns the height. The return parameter documents that height is in meters.

View more solutions

Comments

[oxbow_lakes]

They're only self-explanatory if you say "this is a property setter". Otherwise a client of the API has no idea whatsoever what is actually happening inside the methods

[plankalkul]

Who said anything about self-explanatory?

[Jonik]

Can you fix the typo? "@return part for setters"

[oxbow_lakes]

That's fine - but that requires users of your API to know that, had there been any side effects, they would have been documented !

[mqp]

Without documentation, any caller would assume that a method called setX sets X. It follows that if setX actually sets X, without doing anything else important, then you don't need documentation.

[Steve Kuo]

What if getFirstName() returns null? Where would that be documented?

[Michael Borgwardt]

How about security.getFinalMaturity()? Not all property names have an immediately understandable meaning. Would you want to be fired for not knowing what that means?

[oxbow_lakes]

What if the method is implemented by swizzling? How are you supposed to know that unless it's been clearly documented? How are you supposed to know it is a standard setter unless the doc says it is?

[oxbow_lakes]

That's great! Now does this company CrudTech, whose API I am coding against, follow your convention, or does it follow someone else's on this thread? Hmmmm

[David Berneda]

akf, I was thinking exactly that after posting :) I guess I'll add it to my response.

[theblackips]

I think I like this answer best as there is no redundancy and things like cobertura won't make it look like the coder is slacking with his/her commenting

[Brett]

There's a difference between commenting and documenting.

[Fostah]

There's also a typo in salary()'s comment. It's not JavaDoc comment.

[Fostah]

I agree that it is the best approach to commenting accessors.

[João Marcus]

There is no point in writing "sets the price" in the setPrice doc if the method just sets the value for the price property, but if it also e.g. updates the totalPrice property and recalculates the tax, such behaviour should obviously be documented.

[oxbow_lakes]

No; but "JavaBean property setter" is perfectly descriptive

[matbrgz]

Very true. Exactly therefore is why I do not comment getters and setters. They should be self explanatory, and adding a comment indicates that the code isn't self explanatory.

[matbrgz]

get/set should in my opinion be reserved for getters and setters. Database lookups should be named like "lookupPerson" or so.

[theblackips]

My sentiments exactly, the worst are the domain specific models where only a domain expert knows what the heck the property means.

[littlebroccoli]

You seem to be asking for the documentation to state "This does what you expect." Which is a bit like writing "Caution: HOT" on a cup of coffee. In a perfect world, there would be no need to ever say such things.

[oxbow_lakes]

Yes - having used APIs where methods called things like setX had side-effects other than the expected, I can indeed state with confidence that this is not a perfect world.

[Gad]

while I agree with you, I think that one must make sure that the function comments are not making out for a bad chosen, non-explicit function name.

[Hakanai]

I think the 's' in 'salary' should be in lowercase here, taking into account how these render in the Javadoc output (they are not at the start of a sentence. This Javadoc glitch is a pet peeve of mine.)

[Hakanai]

Another valid answer along these lines might be "designs with getters and setters do not properly understand the notion of encapsulation" :)

[Hakanai]

No good - people don't read the field comments. Javadoc doesn't even generate the private documentation by default, and IDEs don't show you the field documentation when you use quick help on a method call.

[akf]

people don't read the field comments unless they need to. Once there is a need, the more information the better.

[Lyle]

Adding noise to your code for the sake of silencing overly pedantic warning from our tools feels wrong to me. If it doesn't add value to a programmer, then the right solution would be to turn down/fix the verbosity of the tools and/or ease up on how much we care about jumping through hoops so that the tools reward us. Analysis tools are supposed to help us and save effort, not create more senseless tasks for us.

[Laurent Pireyn]

@Trejkaz: Not true, because accessor methods allow for read-only or write-only properties, and for polymorphism (and so wrapping, proxying, and so on).

[Hakanai]

They may allow for those things, but often a builder pattern can replace setters (less mutable) or a visitor pattern can replace getters (more flexible.)

[Vinoth Kumar C M]

Even if it is useful what would you do for getFoo(). Will you copy the same comment for the getFoo() as well ?

[Michael Borgwardt]

@cmv: Obviously the "param" part would not be copied. I'm undecided whether the value of having the information attached to both accessors directly justifies duplicating the information. Probably yes. Even better would be a way to attach one comment to both; I believe this is available in Project Lombok.

[Paŭlo Ebermann]

An effect of this is that now the alphabetic method summary contains no text at all for these methods. This might be good or bad depending on point of view.

[Zordid]

but if you don't document "stupid" getters and setters (it's what I prefer, too!) - how do you get rid of Eclipse warnings on missing javadoc? I don't want to clutter my workspace with warnings like that, but I also don't want that warning to be disabled for all other methods...

[Jarek Przygódzki]

The problem with this approach is that Javadoc doesn't generate the private documentation by default! In that case reference tag {@see #salary} is invalid in generated documentation.

[Sathya Babu Ram k]

@VinothKumar: maybe it would be nicer to simply explain the property in the getter (as in "Foo is the adjustment factor used in the Bar-calculation.") and the effects of changing the value in the setter (or whether it is necessary or not to initialize that value -- in the answer's example it is not necessary to initialize Foo since "it has a default value depending on the Baz type").

[Michael Scheper]

I certainly like and use the builder pattern, but there's so much support for POJOs (e.g. in Hibernate) that getters and setters still have their very prominent place, for better or for worse. It's the daggiest thing about Java, IMHO, and after writing repetitive JavaDocs for well over a decade, I'm about ready to subscribe to @sleske's advice.

[Michael Scheper]

I'm a big supporter of JavaDocs, but also a big supporter of self-documenting code. So for the setter at least, I'd do something like public void setSalary(float aud) (or more realistically, public void setSalary(BigDecimal aud)). Better yet, the property ought to be of type abstract class CurrencyAmount, which in turn has the properties java.util.Currency currency and java.math.BigDecimal amount. Most developers I've worked with are terribly lazy with JavaDoc, but enforcing API like this makes this less of a problem.

[AlexWien]

If the unit is a SI unit like meters / seconds, there is less need to document, if it is not an Si unit then it must be documented or better named to include the non standard unit, e.g heightFeet

[Michael Scheper]

I agree. And then I realised, why write all this boilerplate anyhow? See my answer about Project Lombok.

[Jackson]

+1 for "obscure names that only investment bankers, biochemists or quantum physicists understand"

[Matt Vukas]

@Lyle That may be true, but I feel like there's almost always something useful that a programmer can say which describes a getter/setter better than just the method signature alone. Yes, there are cases where a programmer is lazy and just repeats the method signature in the comment, but I think that generally speaking, it is a helpful behavior to force.