gson supports records

I like using records in Java and getting rid of the boilerplate on my immutable objects. I also like using gson (google’s serialization/deserialization library) for my lightweight JSON needs. All I need is an annotation to use a different field type. (For complex parsing I use Jackson)

Suppose I want parse this

var json = """

For example before records, I would write:

class Name {
	 private String first;
	 private String last;
     @SerializedName("birth_year") int birthYear;

     // boilerplate methods

I needed to use my IDE to generate a constructor, getters, and a toString(). And additionally, equals/hashCode if needed for that class.

In Java 17, I switched to records. I didn’t realize I could use them with gson annotations until recently. Turns out all I need to write is:

record Name(String first, 
   String last, 
   @SerializedName("birth_year") int birthYear) { }

The parsing code remains the same

var gson = new Gson();
var name = gson.fromJson(json, Name.class);

sealed classes, records and enums

Sometimes we write something for the book, only to realize it shouldn’t be in the book. This post is an example. Do you think this legal?

public sealed interface MySealedClass permits MyRecord, MyEnum {

record MyRecord() implements MySealedClass {}

enum MyEnum implements MySealedClass { VALUE }

It is! A sealed interface can be implemented by any Java type. Which means records and enums are included.

Using a Local Record for Readability

Sometimes there is something we really want to put in our cert book that doesn’t fit. You can only have so many case studies/real world scenarios after all. I rescued this from the trash heap and made this blog post./

Suppose we want to count the number of names that have both a first and last name beginning with E. We want to call the method as:

var names = List.of("Sarah Smith", "Eve Edwards");

We can read this with streams:

public long countNamesStartingWithEs(List<String> names) {
      .map(t -> t.split(" "))
      .filter(a -> a[0].startsWith("E"))
      .filter(a -> a[1].startsWith("E"))

The code works. However, it has array positions sprinkled throughout. This requires you to keep remembering the format. We can refactor using a local record:

public long countNamesStartingWithEsWithRecords(List<String> names) {
   record Name(String first, String last) {}

      .map(t -> t.split(" "))
      .map(t -> new Name(t[0], t[1]))
      .filter(n -> n.first().startsWith("E"))
      .filter(n -> n.last().startsWith("E"))

The record lets us give the fields names so we can reference them as n.first() and n.last(). The only place that has to know about the order is the call record constructor. For such a simple example, the approaches are similar. For longer and more complex pipelines, the local record approach can make the code easier to read.