How to avoid checking for null values in method chaining? [duplicate]

Question

No, it is generally not good practice in Java to catch a NPE instead of null-checking your references.

You can use Optional for this kind of thing if you prefer:

if (Optional.ofNullable(country)
            .map(Country::getCity)
            .map(City::getSchool)
            .map(School::getStudent)
            .isPresent()) {
    isValid = true;
}

or simply

boolean isValid = Optional.ofNullable(country)
                          .map(Country::getCity)
                          .map(City::getSchool)
                          .map(School::getStudent)
                          .isPresent();

if that is all that isValid is supposed to be checking.

You could use Optional here, but it creates one Optional object at each step.

boolean isValid = Optional.ofNullable(country)
    .map(country -> country.getCity()) //Or use method reference Country::getCity
    .map(city -> city.getSchool())
    .map(school -> school.getStudent())
    .map(student -> true)
    .orElse(false);

//OR
boolean isValid = Optional.ofNullable(country)
                      .map(..)
                      ....
                      .isPresent();

The object-oriented approach is to put the isValid method in Country and the other classes. It does not reduce the amount of null checks, but each method only has one and you don't repeat them.

public boolean isValid() {
  return city != null && city.isValid();
}

This has the assumption that validation is the same everywhere your Country is used, but typically that is the case. If not, the method should be named hasStudent(), but this is less general and you run the risk of duplicating the whole School interface in Country. For example, in another place you may need hasTeacher() or hasCourse().

Another approach is to use null objects:

public class Country {
  public static final Country NO_COUNTRY = new Country();

  private City city = City.NO_CITY;

  // etc.
}

I'm not sure it is preferable is this case (strictly you would need a sub class to override all modification methods), the Java 8 way would be to go with Optional as method in the other answers, but I would suggest to embrace it more fully:

private Optional<City> city = Optional.ofNullable(city);

public Optional<City> getCity() {
   return city;
}

Both for null objects and Nullable only work if you always use them instead of null (notice the field initialization), otherwise you still need the null checks. So this option avoid null, but you code becomes more verbose to reduced null checks in other places.

Of course, the correct design may be to use Collections where possible (instead of Optional). A Country has a set of City, City has a set of Schools, which has set of students, etc.

As alternative to other fine usage of Optional, we could also use a utility method with a Supplier<Object> var-args as parameter.
It makes sense as we don't have many nested levels in the object to check but many fields to check.
Besides, it may easily be modified to log/handle something as a null is detected.

    boolean isValid = isValid(() -> address, // first level
                              () -> address.getCity(),   // second level
                              () -> address.getCountry(),// second level
                              () -> address.getStreet(), // second level
                              () -> address.getZip(),    // second level
                              () -> address.getCountry() // third level
                                           .getISO()


@SafeVarargs
public static boolean isValid(Supplier<Object>... suppliers) {
    for (Supplier<Object> supplier : suppliers) {
        if (Objects.isNull(supplier.get())) {
            // log, handle specific thing if required
            return false;
        }
    }
    return true;
}

---

Suppose you would like to add some traces, you could so write :

boolean isValid = isValid(  Arrays.asList("address", "city", "country",
                                          "street", "zip", "Country ISO"),
                            () -> address, // first level
                            () -> address.getCity(),   // second level
                            () -> address.getCountry(),// second level
                            () -> address.getStreet(), // second level
                            () -> address.getZip(),    // second level
                            () -> address.getCountry() // third level
                                         .getISO()
                         );


@SafeVarargs
public static boolean isValid(List<String> fieldNames, Supplier<Object>... suppliers) {
    if (fieldNames.size() != suppliers.length){
         throw new IllegalArgumentException("...");
    }
    for (int i = 0; i < suppliers.length; i++) {
        if (Objects.isNull(suppliers.get(i).get())) {
            LOGGER.info( fieldNames.get(i) + " is null");
            return false;
        }
    }
    return true;
}

Java doesn't have "null-safe" operations, like, for example Kotlin's null safety

You can either:

• catch the NPE and ignore it
• check all the references manually
• use Optional as per the other answers
• use some sort of tooling like XLST

Otherwise if you have control over the domain objects, you can redesign your classes so that the information you need is available from the top level object (make Country class do all the null checking...)