Caching and notifying preferences

[michael2402]

Currently, there are many MapFrame.repaint() calls whenever properties change.

Part of this is because colors and other values are cached by each component seperately.

I added a new infrastructure that allows you to get a property and easily cache that result.

To get a property use:

new IntegerProperty("my.property", "123").get()

To cache the result, use:

prop = new IntegerProperty("my.property", "123").cached(); // registers 123 as default value
prop.get(); // lazy get and convert to int
prop.get(); // get cached integer object

I added a new preference listener that listens to changes of one specific key to make this efficient.

[Don-vip]

Patch does not apply anymore, can you please update it?

[Don-vip]

    public static class ValueChangeEvent<T> {
        private final PreferenceChangeEvent base;

        private final AbstractProperty<T> source;

        ValueChangeEvent(PreferenceChangeEvent base, AbstractProperty<T> source) {
            this.base = base;
            this.source = source;
        }

base is not used?

[Don-vip]

Why is the timeout commented in JOSMTestRules?

           //  statement = new FailOnTimeoutStatement(statement, timeout);

[Don-vip]

In 10824/josm:

see #13309 - Caching and notifying preferences (patch by michael2402) - gsoc-core

[michael2402]

Thanks for finding that timeout. No, it should not be commented. I did this for debugging.

[Don-vip]

ok this is fixed, I didn't apply this part.

[Don-vip]

this is not, this patch cause 74 unit test failures: jenkins/job/JOSM/1725/testReport/

[Don-vip]

In 10839/josm:

see #13309 - remove deprecated color stuff not used outside of JOSM

[Don-vip]

In 10853/josm:

see #13309 - fix most of deprecation warnings

[Don-vip]

it seems the exception causing all these test failures is

java.lang.NullPointerException
	at org.openstreetmap.josm.data.preferences.AbstractToStringProperty.getAsString(AbstractToStringProperty.java:131)
	at org.openstreetmap.josm.data.preferences.AbstractToStringProperty.get(AbstractToStringProperty.java:92)
	at org.openstreetmap.josm.data.preferences.BooleanProperty.get(BooleanProperty.java:21)
	at org.openstreetmap.josm.tools.date.DateUtils.getDateTimeFormat(DateUtils.java:311)
	at org.openstreetmap.josm.gui.dialogs.NotesDialog$NoteRenderer.<init>(NotesDialog.java:209)
	at org.openstreetmap.josm.gui.dialogs.NotesDialog$NoteRenderer.<init>(NotesDialog.java:206)
	at org.openstreetmap.josm.gui.dialogs.NotesDialog.buildDialog(NotesDialog.java:85)
	at org.openstreetmap.josm.gui.dialogs.NotesDialog.<init>(NotesDialog.java:78)
	at org.openstreetmap.josm.gui.MapFrame.<init>(MapFrame.java:264)
	at org.openstreetmap.josm.gui.MainPanel.createNewMapFrame(MainPanel.java:88)
	at org.openstreetmap.josm.gui.MainPanel.updateContent(MainPanel.java:63)
	at org.openstreetmap.josm.gui.MainPanel$1.beforeFirstLayerAdded(MainPanel.java:157)
	at org.openstreetmap.josm.gui.layer.MainLayerManager.realAddLayer(MainLayerManager.java:264)
	at org.openstreetmap.josm.gui.layer.LayerManager.lambda$addLayer$0(LayerManager.java:180)
	at org.openstreetmap.josm.gui.util.GuiHelper.runInEDTAndWaitWithException(GuiHelper.java:138)
	at org.openstreetmap.josm.gui.layer.LayerManager.addLayer(LayerManager.java:180)
	at org.openstreetmap.josm.JOSMFixture.setupGUI(JOSMFixture.java:154)
	at org.openstreetmap.josm.JOSMFixture.access$000(JOSMFixture.java:28)
	at org.openstreetmap.josm.JOSMFixture$1.run(JOSMFixture.java:128)

[michael2402]

The problem is that some class loads DateUtils before Main.pref is instantiated.

I suggest fixing this and adding a dependency on preferences for that test. I'll find the test that causes this and post a patch for it.

An alternative to make our life easier would be to make Main.pref a final field instantiated with an empty instance. We can then simply load it during program startup. This would also make early application startup much easier, since we currently have many classes that may only be loaded after Main.pref is set. We should make those classes listen to preferences updates though, but this is why I created this patch.