JavaFX Displaying properties in controls

Question

I'm working through myself Oracle's JavaFX tutorials. After doing Swing for many years (a long time ago) I'm fascinated by the new smart features, incl. properties. I was surprised to see that these examples (e.g: https://docs.oracle.com/javafx/2/ui_controls/table-view.htm) don't use them in a way what I believe to be "proper".

The example creates a Person class with properties as fields:

public static class Person {
    private final SimpleStringProperty firstName;
    ...

But the getters are not for the properties, but for their values

    public String getFirstName() {
        return firstName.get();
    }

so when it binds those to TableCells in columns, it wraps them in a new property:

    emailCol.setCellValueFactory(
            new PropertyValueFactory<Person, String>("firstName"));

This seems convoluted to me, and misses the real advantage of event propagation, by not simply using this:

    firstNameCol.setCellValueFactory( celldata -> 
        celldata.getValue().firstNameProperty());

My question: Is there a reason why this example doesn't expose and use the bean's properties in the control directly? Am I missing something here?

NB: I did change the code this way, and the example worked much nicer: updates in the Person entity by an other control were propagated immediately, w/o calling table.refresh() e.g.

going a bit further: if you need refresh(), something is wrong with your setup — kleopatra, Feb 28 '18 at 03:42
The [tutorial for `TableView`](https://docs.oracle.com/javase/8/javafx/user-interface-tutorial/table-view.htm#CJAGAAEE) actually fails to define the model class (`Person`) correctly. The [properties and bindings tutorial](http://www.oracle.com/pls/topic/lookup?ctx=javase80&id=JFXBD107), or the [Makery tutorial](http://code.makery.ch/library/javafx-8-tutorial/part2/) show the correct approach. — James_D, Feb 28 '18 at 03:46
@James_D yeah, it's a bug in the tutorial https://bugs.openjdk.java.net/browse/JDK-8089693 even though the maintainers don't think so ;) — kleopatra, Feb 28 '18 at 08:35

James_D · Accepted Answer · 2018-02-28T03:55:23.483

First, note that if you follow the expected pattern:

public class Person {

    private final StringProperty firstName = new SimpleStringProperty();

    public StringProperty firstNameProperty() {
        return firstName ;
    }

    public final String getFirstName() {
        return firstNameProperty().get();
    }

    public final void setFirstName(String firstName) {
        firstNameProperty().set(firstName);
    }
}

then either version of your code will work without calling table.refresh(). This was the intended use of PropertyValueFactory, as is fairly clear from the documentation.

You are correct, though, that the lambda expression is a better approach than the PropertyValueFactory. Besides the reasons you cite, there are other major advantages to using the lambda expression over PropertyValueFactory. First, and most importantly, the PropertyValueFactory simply takes the name of the property as a String, meaning that there is no compile-time check for it. So if you were to misspell the name of the property:

firstNameCol.setCellValueFactory(new PropertyValueFactory<>("firstname"));

this would compile just fine, and you would simply end up with blank cells in the column. This can be quite hard to debug (as is evidenced by the number of questions on this site asking for help with this kind of bug: e.g. Javafx PropertyValueFactory not populating Tableview).

Secondly, the PropertyValueFactory works by reflection, which is much slower than a lambda expression. This can lead to measurable performance differences, for example in sorting a table with a large amount of data.

The reason PropertyValueFactory was introduced is basically historical. Prior to Java 8, there were of course no lambda expressions, so the minimal implementation of the cell factory without this convenience class was via an anonymous inner class:

firstNameCol.setCellValueFactory(new Callback<TableColumn.CellDataFeatures<Person, String>, ObservableValue<String>>() {
    @Override
    public ObservableValue<String> call(TableColumn.CellDataFeatures<Person, String> cellData) {
        return cellData.getValue().firstNameProperty();
    }
});

Since that code is really quite hideous, the JavaFX team introduced the PropertyValueFactory simply as a way to make the API easier to use.

When using Java 8 and later, PropertyValueFactory should really be considered a legacy class, and the lambda expression should be preferred. Of course, documentation that predates Java 8 also still exists (in fact, you explicitly link documentation from JavaFX 2 - though the most recent version still hasn't been updated), and there are - to be frank - far too many other writers who copy that style without properly thinking it through. There is probably a good case to be made for deprecating the PropertyValueFactory class entirely.

(So: TL;DR: no, you're not missing anything.)

JavaFX Displaying properties in controls

1 Answers1

Linked