21

I've been using Guava's ImmutableMap with a Builder for awhile without giving much thought to how/why it works. I've been using the builder in the way described by the Javadoc:

ImmutableMap<String, Integer> WORD_TO_INT =
          new ImmutableMap.Builder<String, Integer>()
              .put("one", 1)
              .put("two", 2)
              .put("three", 3)
              .build();

I've also seen examples elsewhere on SO of this syntax:

ImmutableMap<String,Integer> myMap = ImmutableMap.<String, Integer>builder()
    .put("one", 1) 
    .put("two", 2) 
    .put("three", 3) 
    .build();

Both seem to produce the exact same result.

A quick glance at the ImmutableMap source shows that the static builder() call in the 2nd example returns: new Builder<K, V>()

There doesn't appear to be any functional difference between the two methods.

Is there any difference under the hood? Any reason to prefer one or the other?

EDIT: Adding bytecode difference.

The bytecode generated is almost identical except for one line:

Method 1(Builder<K,V>):

static {};
     0  new com.google.common.collect.ImmutableMap$Builder [12]
     3  dup
     4  invokespecial com.google.common.collect.ImmutableMap$Builder() [14]
     7  ldc <String "key1"> [17]
     ........

Method 2:(<K,V>builder())

static {};
     0  invokestatic com.google.common.collect.ImmutableMap.builder() : com.google.common.collect.ImmutableMap$Builder [12]
     3  ldc <String "key1"> [18]
     ........

After that they're pretty much the same, the first line being different makes sense given the source differences, but I'm not sure why the first one calls dup and the second doesn't.

2 Answers 2

21

No, they are identical. The reason for the first syntax is that it makes a nice one-liner.

The reason for the second syntax is the you might have a loop or other flow control, you may need to pass the Builder around.

For example:

public static void main(final String[] args) throws Exception {
    final ImmutableList.Builder<Integer> lb = ImmutableList.builder();
    for (final String arg : args) {
        lb.add(Integer.parseInt(arg));
    }
}

And before you say it, yes a FluentIterable would probably be better in this case - but it's just an example.

Sign up to request clarification or add additional context in comments.

5 Comments

Any idea why the dup shows up in the bytecode for the first one? I don't know enough about bytecode to know if that's typical or not.
@MagicMan if you look at the source code you will see that the methods are different - I'm not surprised that the bytecode is different. As to why it duplicates the value, I'm not sure.
Ahh, as long as there isn't any big difference I won't worry. Example makes it clear how I should be using it, thanks!
@MagicMan, you didn't provide full bytecode listing, but I believe that there is something like invokespecial ...ImmutableMap$Builder.<init>() after dup in the first example. This is exactly why dup is needed. In Java construction is two-phase process: first, new memory is allocated via new, and then a constructor method (<init>) is called on that memory. So duped stack element is consumed by the constructor.
@VladimirMatveev Ahh, sorry, just edited and added that next line. You are correct, it was indeed invokespecial....
0

As stated in the javadoc, The generated builder is equivalent to the builder created by the ImmutableMap.Builder constructor.

It is a good practice to write such a method when applying the Builder pattern. This offers a fluent way to use it and can be considered as a "syntactic sugar".

Comments

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.