Question about Effective Java

[Novice]

I'm trying to understand the example on page 14 of Effective Java by Joshua
Bloch (Second Edition). I hope someone here can help.

Here's the code from the book for the sake of those who don't have the book
(or the same edition of the book):
=======================================================================
//Builder Pattern
public class NutritionFacts {
private final int servingSize;
private final int servings;
private final int calories;
private final int fat;
private final int sodium;
private final int carbohydrate;

public static class Builder {
//Required parameters
private final int servingSize;
private final int servings;

//Option parameters - initialized to default values
private int calories = 0;
private int fat = 0;
private int carbohydrate = 0;
private int sodium = 0;

public Builder(int servingSize, int servings) {
this.servingSize = servingSize;
this.servings = servings;
}

public Builder calories(int val) {
this.calories = val;
return this;
}

public Builder fat(int val) {
this.fat = val;
return this;
}

public Builder carbohydrate(int val) {
this.carbohydrate = val;
return this;
}

public Builder sodium(int val) {
this.sodium = val;
return this;
}

public NutritionFacts build() {
return new NutritionFacts(this);
}
}
private NutritionFacts(Builder builder) {
this.servingSize = builder.servingSize;
this.servings = builder.servings;
this.calories = builder.calories;
this.fat = builder.fat;
this.sodium = builder.sodium;
this.carbohydrate = builder.carbohydrate;

}

}=======================================================================

I should start by saying I have successfully imitated the example in my own
code. I'm just not entirely sure what it is doing.

Bloch clearly intends this technique as a way to avoid the confusion of
methods where you simply have a long list of parameters and could easily
enter parameters out of order. The code on page 15 shows how much clearer
things are when you use his technique; all of the optional parameters can
have clear names so that you know what each one is. I really like this; I
think it's MUCH better than simply listing a bunch of different integers
and hoping you assign each one to the right variable. In a nutshell, I get
what he is doing but I'm just not sure how this code works.

Questions:
1. I understand that a NutritionFacts could conceivably be just a
servingSize and a number of servings. All of the other values are optional
on an individual basis: any one or more of them could be there but don't
need to be. Any value that is omitted takes a default (of zero in this
example). But why are the methods calories(), fat(), carbohydrate() and
sodium() each returning Builder? They aren't constructors and they only
deal with one data value, fat or calories or whatever, but they apparently
return a complete Builder.

I'm also having trouble understanding some of the compiler warnings given
to me by Eclipse (3.7.1) when I actually code the class the way he has it:
- I get a "not used" warning for all of the private final variables in
NutritionClass (Eclipse offers to delete the variables OR create gettsrs
and setters OR "add @SuppressWarnings - unused" for each of them. I'm not
fond of any of these options but what is the best thing to do and why? Why
does Eclipse think they are unused when the NutritionFacts constructor is
clearly assigning values to them?
- I get "access to enclosing constructor NutritionFacts
(NutritionFacts.Builder) is emulated by a synthetic accessor method" in the
return statement of the build() method (Eclipse offers to remove the
"private" from the constructor for NutritionFacts OR to add
"@SuppressWarnings - synthetic access" to NutritionFacts but I'm not sure
which is better and why.
- I get "read access to enclosing field NutritionFacts.Builder.servingSize
is emulated by a synthetic accessor method" (Eclipse offers to remove the
"private" from the variables in Builder OR to add "@SuppressWarnings -
synthetic access" to NutritionFacts but I'm not sure which is better and
why.

 

[markspace]

I'm trying to understand the example on page 14 of Effective Java by Joshua

Questions:
1. I understand that a NutritionFacts could conceivably be just a
servingSize and a number of servings. All of the other values are optional
on an individual basis: any one or more of them could be there but don't
need to be. Any value that is omitted takes a default (of zero in this
example). But why are the methods calories(), fat(), carbohydrate() and
sodium() each returning Builder?

So you can chain calls, like the example at the top of page 15.

They aren't constructors and they only
deal with one data value, fat or calories or whatever, but they apparently
return a complete Builder.

They return a Builder object (which might be the same thing as saying "a
complete Builder"). Constructors don't have a return type explicitly.
Learn Java syntax.

public class Builder {
public Builder() {} // constructor
public Builder doIt() {} // method
}

I'm also having trouble understanding some of the compiler warnings given
to me by Eclipse (3.7.1) when I actually code the class the way he has it:
- I get a "not used" warning for all of the private final variables in

Right, it's just a dummy example. If you add getters to NutritionFacts,
those warnings will go away.

NutritionClass (Eclipse offers to delete the variables OR create gettsrs
and setters OR "add @SuppressWarnings - unused" for each of them. I'm not
fond of any of these options but what is the best thing to do and why? Why
does Eclipse think they are unused when the NutritionFacts constructor is
clearly assigning values to them?

It's a Java thing. Unused means can't be read, not is not initialized.
Things that are initialized and then never used merit a warning message.

 

[Novice]

So you can chain calls, like the example at the top of page 15.

I understand that chaining is what we'd like to be able to do. I'm just
not clear on why "this", which from the method signature is apparently a
Builder, is being returned by those methods. It _looks_ like each of them
is returning an entire Builder even though each only contributed one
"fact" to the Builder.

They return a Builder object (which might be the same thing as saying
"a complete Builder"). Constructors don't have a return type
explicitly. Learn Java syntax.

public class Builder {
public Builder() {} // constructor
public Builder doIt() {} // method
}

Right; I understand that already. The phrasing of my question is probably
weird because I can't quite understand what is happening in the example.

Right, it's just a dummy example. If you add getters to
NutritionFacts, those warnings will go away.

But the body of the NutritionFacts constructor is already setting values
for the variables. I proved that to myself by changing the name of one of
the NutritionFacts final variables and causing a compiler error in the
NutritionFacts constructor. Why would getters be needed if the
constructor is already setting values for the variables?

It's a Java thing. Unused means can't be read, not is not
initialized.

Things that are initialized and then never used merit a warning
message.

So if I only initialize it and never read it after that, I can still get
an "unused" message? If that is right, it explains why the NutritionFacts
constructor code doesn't prevent the message.

I could have sworn that I've made that warning go away in the past by
simply doing an assignment statement to give the variable a value, even
if I never touched it again later. Maybe I'm misremembering that....

 

[Novice]

Don't think in terms of "return a complete Builder". Java programs do
not, and cannot, pass objects around as method arguments or results.
sodium really returns a Builder reference. A reference can either be
null or be a pointer to an object of appropriate class.

Inside sodium, "this" is a pointer to the current object. Returning it
allows chaining of several method calls.

I'll have to mull that over for a bit and consider the implications.

At first glance it seems simple enough; in any other class, "this" refers
to an instance of the class that contains the "this", just like we see
here. But I'm not used to seeing several methods in a class all return
"this". I'm also not used to named classes nested within classes. (I'm
familiar with anonymous inner classes like window listeners and with
putting additional named classes within the same source file in a
sequential fashion but putting a named class within another is still
feels a little odd.)

I'll post back if I can't make my peace with this slightly exotic
technique ;-)

Thanks for your help.

 

[markspace]

I understand that chaining is what we'd like to be able to do. I'm just
not clear on why "this", which from the method signature is apparently a
Builder, is being returned by those methods. It _looks_ like each of them
is returning an entire Builder even though each only contributed one
"fact" to the Builder.

Patricia had an insight that I missed, which was that only a reference,
a pointer to the object, gets returned, and indeed that's all Java uses
is pointers, not "complete objects".

So there's only one Builder, which sits on the heap, and the return
statement keeps passing a reference to that object back to the code. I
usually don't make a distinction because the practical difference is
small. But it's there, and can sometimes be helpful when you analyze code.

Also, "this" doesn't mean of the class Builder. It could be a subclass;
but inheritance means that you can't tell, and usually shouldn't care.

So if I only initialize it and never read it after that, I can still get
an "unused" message? If that is right, it explains why the NutritionFacts
constructor code doesn't prevent the message.

Yup, unused and can't be used, because they're private. If you also
made them public, then some other class might use them, even if none
actually do. That would also make the warning go away.

 

[Eric Sosman]

I'll have to mull that over for a bit and consider the implications.

At first glance it seems simple enough; in any other class, "this" refers
to an instance of the class that contains the "this", just like we see
here. But I'm not used to seeing several methods in a class all return
"this".

Analogies are never perfect, but often helpful. So, imagine a
group of friends ordering pizza: They've got a slip of paper listing
possible toppings, and they're going to put check marks next to the
items they want. "Pepperoni," says George, "gotta have pepperoni,"
making his mark on the paper and passing it to Gina. "Mushrooms
make my breath fresher," she says, adding her mark and passing the
paper along to Gail. She opts for extra cheese and hands the paper
to Guy, who's worried about his weight and says "Sounds good, let's
just go with these three." He passes the paper over the counter to
Guido, who makes them a pluperfectly personalized pizza.

Now, what do the friends pass among themselves during this
colloquy? Do they pass the pizza? No, it hasn't even been made
yet. Do they pass their chosen ingredients? No, George doesn't
have a handful of pepperoni, nor does Gina have mushrooms nor Gail
extra cheese, and Guy doesn't have plenty of nothing. What they
pass is the slip of paper, the specification of the pizza they would
like Guido to build for them.

... and that's very much like what's going on with Builder and
NutritionFacts. The friends "make marks on" a Builder, then hand
the Builder to Guido, who makes them a NutritionFacts built to its
specification.

That's the idea; now to the returning of "this". It's just a
convenience, really. Imagine that all those Builder methods were
"void", returning nothing at all. Then you'd write

NutritionFacts.Builder b = new NutritionFacts.Builder(300, 4);
b.calories(950);
b.sodium(800);
b.carbohydrate(1025);
NutritionFacts facts = b.build();

That is, you'd use the variable "b" to hold a reference to the
Builder while you "make your marks" on it. Eventually, you call
its build() method to "hand it to Guido" and get a pizza -- er,
that is, a NutritionFacts -- made to order. This would work just
fine, and there's nothing fundamentally wrong with it.

But if each of the Builder method returns a reference to the
Builder it has just "marked" things can be written a little more
compactly:

NutritionFacts facts =
new NutritionFacts.Builder(300, 4)
.calories(950)
.sodium(800)
.carbohydrate(1025)
.build();

Observe that you no longer need the "b" variable: The "new" creates
a Builder and produces a reference to it, the calories() method
modifies that Builder and returns a reference that the sodium()
method can use, ..., until eventually the build() method takes all
the information stored in the Builder and uses it to generate a
Pizza -- er, that is, a NutritionFacts. Since there's no "b" to
refer to the Builder, it becomes eligible for garbage collection
as soon as its fulfilled its purpose, namely, specifying what's
wanted in the new NutritionFacts.

In short, the returning of "this" enables a convenience, but
is not truly fundamental to the pattern of using a builder of some
kind to hold the parameters for a hidden constructor.

I'm also not used to named classes nested within classes. (I'm
familiar with anonymous inner classes like window listeners and with
putting additional named classes within the same source file in a
sequential fashion but putting a named class within another is still
feels a little odd.)

You'll get used to it. The pattern works even if the builder
class is completely separate from the built class -- consider, for
example, StringBuilder and String. (And note how StringBuilder's
append() method, for instance, returns a reference to the same
StringBuilder that the append() modified.)

I'll post back if I can't make my peace with this slightly exotic
technique ;-)

You'll get used to it. (Is there an echo in here?)

 

[John B. Matthews]

I'm trying to understand the example on page 14 of Effective Java by Joshua
Bloch (Second Edition). I hope someone here can help.

Here's the code from the book for the sake of those who don't have the book
(or the same edition of the book):

Excerpted here: <http://drdobbs.com/jvm/208403883?pgno=2>.

[...]

 

[Lew]

Get used to it.

Have you considered Googling for "builder pattern"? You might be surprised by
what just a few minutes of reading can reveal.

Analogies are never perfect, but often helpful. So, imagine a
group of friends ordering pizza: They've got a slip of paper listing
possible toppings, and they're going to put check marks next to the
items they want. "Pepperoni," says George, "gotta have pepperoni,"
making his mark on the paper and passing it to Gina. "Mushrooms
make my breath fresher," she says, adding her mark and passing the
paper along to Gail. She opts for extra cheese and hands the paper
to Guy, who's worried about his weight and says "Sounds good, let's
just go with these three." He passes the paper over the counter to
Guido, who makes them a pluperfectly personalized pizza.

Now, what do the friends pass among themselves during this
colloquy? Do they pass the pizza? No, it hasn't even been made
yet. Do they pass their chosen ingredients? No, George doesn't
have a handful of pepperoni, nor does Gina have mushrooms nor Gail
extra cheese, and Guy doesn't have plenty of nothing. What they
pass is the slip of paper, the specification of the pizza they would
like Guido to build for them.

... and that's very much like what's going on with Builder and
NutritionFacts. The friends "make marks on" a Builder, then hand
the Builder to Guido, who makes them a NutritionFacts built to its
specification.

That's the idea; now to the returning of "this". It's just a
convenience, really. Imagine that all those Builder methods were
"void", returning nothing at all. Then you'd write

NutritionFacts.Builder b = new NutritionFacts.Builder(300, 4);
b.calories(950);
b.sodium(800);
b.carbohydrate(1025);
NutritionFacts facts = b.build();

That is, you'd use the variable "b" to hold a reference to the
Builder while you "make your marks" on it. Eventually, you call
its build() method to "hand it to Guido" and get a pizza -- er,
that is, a NutritionFacts -- made to order. This would work just
fine, and there's nothing fundamentally wrong with it.

But if each of the Builder method returns a reference to the
Builder it has just "marked" things can be written a little more
compactly:

NutritionFacts facts =
new NutritionFacts.Builder(300, 4)
.calories(950)
.sodium(800)
.carbohydrate(1025)
.build();

Observe that you no longer need the "b" variable: The "new" creates
a Builder and produces a reference to it, the calories() method
modifies that Builder and returns a reference that the sodium()
method can use, ..., until eventually the build() method takes all
the information stored in the Builder and uses it to generate a
Pizza -- er, that is, a NutritionFacts. Since there's no "b" to
refer to the Builder, it becomes eligible for garbage collection
as soon as its fulfilled its purpose, namely, specifying what's
wanted in the new NutritionFacts.

In short, the returning of "this" enables a convenience, but
is not truly fundamental to the pattern of using a builder of some
kind to hold the parameters for a hidden constructor.

Read the Java tutorials.

Then study the Java Language Specification (JLS) about "member types".

The keyword is "scope". The scope of a nested class is defined like any other
class or instance member: 'public', 'protected', package-private or 'private'.
There are also some special rules about access to the enclosing type's members.

It's the same, but anonymous classes don't have a name (hence "anonymous").

Do note that not all nested classes are inner classes; some are static.

"Not familiar", "feels a little odd". Jeez, get over it. You sound like a
whiney little baby.

If you haven't got the ability to learn new stuff, you won't last 10 minutes in
software development. Really, get over it.

You'll get used to it. The pattern works even if the builder
class is completely separate from the built class -- consider, for
example, StringBuilder and String. (And note how StringBuilder's
append() method, for instance, returns a reference to the same
StringBuilder that the append() modified.)

Why don't you first study up on it? The newsgroup is not a help desk, and it
certainly is no substitute for study.

And you call it "exotic" only because you're incompletely trained. Don't make
such judgments given your lack of experience. You are in no position yet to
decide what is "exotic" and what isn't. You need to check your ego at the door.

You'll get used to it. (Is there an echo in here?)

Bottom line: Study, and get used to learning new things.