Javadoc

There are three packages in this API:

Contents

Datasystems API

Loaders and Data Objects

There are a few related object types used in NetBeans to handle creation of structured, high-level data based on files found on a filesystem. In summary, the system loader pool is responsible for scanning files in a directory on disk, weeding out irrelevant files of no interest, and grouping the rest into logical chunks, or just determining what type of data each represents. It does this scanning by asking each registered data loader whether or not the given file(s) should be handled. The first loader to recognize a file takes ownership of it, and creates a matching data object to represent it to the rest of NetBeans.

What is a loader used for?

There are a few reasons why loaders are used in NetBeans, rather than just accessing raw files (and perhaps typing them by extension or MIME type):

Multiple-file loaders

These typically subclass MultiFileLoader. Like the loader used by the Form Editor, they are able to recognize multiple files at once and create a data object from them. All data objects have a primary file which is representative of the data object; e.g. myform.java in the previous example. As well, these multi loaders may have any number of secondary files.

The basic mechanism a multi-file loader uses, is that the loader pool will pass it file objects to recognize in an arbitrary order; the loader may get a primary or secondary file first. Either way, it must recognize that it belongs to a cluster; find the primary file if it got a secondary one; and create a new MultiDataObject containing the supplied file as an entry. When other files in the cluster are passed to the loader, it must create new entries in the same multi data object, to indicate their association.

Single-file loaders

These typically subclass UniFileLoader. A single-file loader is of course simpler, and is likely to be more commonly used (since the majority of file types make sense by themselves). The default implementation makes it straightforward to create a subclass recognizing only certain file extensions.

This kind of loader may be used for, e.g., HTML files which ought to be recognized as such and given a simple data object appropriate to working with HTML (so that opening it by default launches a web browser, and so on).

Note that the standard UniFileLoader is actually a special type of MultiFileLoader (only recognizing one file), so that it actually creates a MultiDataObject when it recognizes its file. Normally you will use a UniFileLoader for most purposes, so the behaviors in MultiDataObject are generally available as defaults. If you had some reason to avoid using this kind of data object, you could of course subclass DataLoader and DataObject directly.

Entries

Entries represent a single file in a MultiDataObject, and thus are commonly used in all loaders (even single-file loaders). Normally MultiDataObject.Entry will not be specially subclassed; most module authors will use FileEntry (a regular entry) and sometimes FileEntry.Numb (a discardable file which should not be slavishly moved about just because the primary file is).

Entries primarily handle operations on the whole file, such as copying or instantiation from template, and provide the ability to insert special behavior hooks into such operations, if desired. For example, a Java source entry might want to rename its package and class name when it was moved. (The easiest way to create "smart" templates such as this is to make the entry a FileEntry.Format.)

Getting a data object

Normally you do not need to explicitly retrieve a data object - the loader pool will create them on demand, present them to the user in the Explorer as nodes, and user actions will trigger use of the data object. However, if you do need to retrieve the data object for a file object you may do so using DataObject.find(...), which will create a new data object for the file if it needs to, otherwise will return the existing one. (Catch DataObjectNotFoundException in case the file is not recognizable to any loader.)

Node Delegates

Data objects have separate, but associated, Nodes which represent them in the Explorer and present a meaningful interface to the user. You may use DataObject.getNodeDelegate() to retrieve the node delegate for a data object, if necessary.

Writing a loader and data object

This section is the most important for most people: after a loader has been correctly written and installed, the system takes care of providing files for it to recognize; constructing nodes for the files; etc.; and most further implementation of a module will be callbacks run in response to some aspect of the objects created here.

Module installation

First of all, you must be able to register your loader in the module so that NetBeans knows about it. The preferred way since version 7.0 of org.openide.loaders module is to assign the loader to appropriate MIME type. If the FileObject of your interest returns for example text/xml+acme from its getMIMEType() method, then you shall register its DataLoader or other configured DataObject.Factory into layer file of your module:

<folder name="Loaders">
  <folder name="text">
    <folder name="xml+acme">
      <folder name="Factories">
        <file name="org-acme-pkg-AcmeLoader.instance">
          <attr name="position" intvalue="100"/>
        </file>
      </folder>
    </folder>
  </folder>
</folder>
Then the AcmeLoader will be consulted everytime the system needs to recognize file with text/xml+acme MIME type. Please refer to separate specification document on information how to register a MIME type declaratively.

The recognition process for any file consults loaders registered in an old, deprecated style first, then it checks folder Loaders/mime/type/Factories for all instances of DataObject.Factory which also includes instances of all DataLoaders. If no appropriate loader is found, the scan continues among loaders registered as Loaders/content/unknown/Factories. At the end the recognition checks standard system loaders. In case of registration of multiple loaders in one folder you can specify their order by using the position attributes. Pay attention to whether this loader is potentially in conflict with other existing or probable loaders; if so, you should specify that it take higher or lower precedence than these others. (This would be necessary, e.g., if your loader recognized everything that another loader did, and used that loader, but also added further special behavior.)

Deciding what to subclass

Your first step in writing the loader is deciding what to subclass. Though you could in principle subclass DataLoader directly, in practice it is easier and better to subclass either MultiFileLoader or UniFileLoader, according to whether your loader needs to cluster files, or just needs to deal with a single primary file.

Handling file recognition

For a single-file loader, handling file recognition is quite straightforward from the point of view of the API: you just need to override UniFileLoader.findPrimaryFile(...) to return its argument if this file is of the type that should be handled by your loader, or null if it is not.

In fact, providing that your loader is of the common sort that just looks for a specific file extension, you do not even need to override this method at all; simply create an appropriate ExtensionList and call UniFileLoader.setExtensions(...) in your loader's constructor.

For a multi-file loader, the situation is slightly more complex, but still NetBeans takes care of most of the work for you. You should implement MultiFileLoader.findPrimaryFile(...) as follows:

You need not worry about which order the primary and secondary files are recognized in, or whether some other loader may have gotten to a file first - the implementation in NetBeans takes care of this for you.

Creating the data object

The most interesting part of the loader is that which actually creates the data object from the underlying file object. You must implement MultiFileLoader.createMultiObject(...) in order to do this. The method will be passed the correct primary file object for you to work with.

To write the data object, subclass MultiDataObject. Your constructor will call the superclass constructor (so the loader will pass in the desired primary object and a reference to itself). You do not need to worry about whether or not to throw DataObjectExistsException in the constructor; it will be thrown automatically by the superclass if necessary.

After that, what to override in the data object is up to you. Other than things mentioned below, you may find it useful to prevent the data object from being renamed or otherwise tampered with, if doing so would make it useless or corrupted in some way; just return false from e.g. MultiDataObject.isRenameAllowed().

Or, if e.g. moves are to be permitted but require special treatment, you may override e.g. MultiDataObject.handleMove(...).

Choosing entry types

For single-file loaders, the default implementation of UniFileLoader.createPrimaryEntry(...) just produces a FileEntry, which is most likely what you want.

For multi-file loaders, you must explicitly select the entry types by implementing MultiFileLoader.createPrimaryEntry(...) and MultiFileLoader.createSecondaryEntry(...).

Typically, the primary entry will be a FileEntry, and will behave normally. The secondary entry might also be a FileEntry, if it makes sense to move the secondary entries along with the primary (i.e., if they are valuable enough to do so, and will not be corrupted); in many cases you will want to use a FileEntry.Numb, which will not be moved along with the primary file, and may just be discarded (for example, this would be useful for compiled *.class files, cached indices, etc.). For such dummy files, you will generally also want to use FileObject.setImportant(...) to prevent the file from being considered by a version control system, for example.

It is possible to specify custom copy/move/etc. behavior for individual files in your data object by subclassing MultiDataObject.Entry (or FileEntry) and providing a non-obvious implementation. If you need custom behavior for the whole data object at once, it is preferable to do so by overriding methods on the data object, as mentioned above.

FileEntry.Format is a convenient entry type to use if you wish to perform substitution of some type of token when creating the file from template. Typically the method FileEntry.Format.createFormat(...) will be implemented to return an instance of MapFormat containing substitution keys and values according to the name and package of the file object; constants used by the module; values of associated system options; the current time and date or user name; etc. For example, the Java data loader uses this entry type with a customized MapFormat permitting it to replace keys such as __NAME__ with the (new) name of the class, or __USER__ with the current user name (as taken from a system option, defaulted from the Java system property).

Loader state and bean info

Data loaders all implicitly extend SharedClassObject, which means that there is only intended to be a single instance of the loader per class, and all associated configuration and properties are stored in a shared state pool. SharedClassObject manages this state implicitly; if you wish to associate any properties with a data loader, you should:
  1. Implement getter and setter methods to call SharedClassObject.getProperty(key) and SharedClassObject.putProperty(key, value, true) (the latter will automatically fire property changes and synchronize for you).
  2. Override readExternal and writeExternal to read and write your property values from the stream. Please always first call the super methods.

You should use the method SharedClassObject.initialize() to set up the shared instance, including both your own properties, and standard ones such as DataLoader.setDisplayName(String), DataLoader.setActions(SystemAction[]), and UniFileLoader.setExtensions(ExtensionList).

Finally, data loaders will be customized by the user as Beans (and persisted using externalization). For this reason, they should have an associated bean info class which should typically specify:

  1. A list of additional bean infos to retrieve (introspect on the superclass).
  2. Display names and hints (and property editors, etc.) for all properties you define.
  3. An icon.

Providing cookies and actions

The most common way in which a new data object type will present its useful features to NetBeans is by adding cookies and actions specific to it. Please see the cookie subsection of this document for information on cookies, and the Actions API for details on how to write actions.

One simple way to add cookie support to a data object is simply to implement the cookie's interface on the data object itself; then it will automatically support the cookie. But doing so for many cookies may be a bad idea, as your data object class will become cluttered; and there is no way to alter the set of cookies provided in this way.

A better technique is to provide cookies explicitly from DataObject.getCookie(...). Assuming that you are subclassing MultiDataObject, you need not override this method yourself, but rather should use MultiDataObject.getCookieSet() in the constructor and add the cookies you want to provide by default. Then it is possible to extend this set later, and to more easily examine its contents.

You may attach some actions to the nodes associated with your data objects. The easiest way to do this is to call DataLoader.setActions(...) in your loader's SharedClassObject.initialize() method, which lets you provide a set of actions appropriate to all data objects created by this loader. Please see the Actions API for an example displaying suggested standard actions to include, and their positioning.

Or, you may wish to selectively attach actions to certain data objects' nodes and not others. If you need to do this, please override DataNode.getActions(boolean) when creating your node delegate; you probably want to call the super method and append any additional actions this particular node should have.

The nodes ought also to have a default action, which will be performed in response to a generic user-initiated event, such as a double-click on the node; this should do something safe and obvious on the node, such as opening it for editing, running it if executable, etc. To do so, your node delegate should override DataNode.getPreferredAction(). If unspecified, NetBeans may still provide a generic default action, such as displaying properties of the object.

The default implementation is only specified in the case of templates, so you may override this. However, if there is a chance this data object might serve as a template, for UI consistency this default action should be preserved; you may check DataObject.isTemplate(), and if true, provide InstantiateAction as the result.

Creating a node delegate

You must create a Node to represent your data object in the Explorer hierarchy, so that the user may interact with it visually. The method DataObject.createNodeDelegate() controls what sort of node should be created. To control the icon, you should use AbstractNode.setIconBaseWithExtension(...) on the newly created DataNode, either in its constructor (if subclassing) or in DataObject.createNodeDelegate().

You have considerable latitude in creating this node; e.g. the Form Editor actually creates a full hierarchy for nodes representing forms, including one subtree representing the Java class object (and its various members), as well as a subtree representing the component structure of the form (as displayed in the Component Inspector).

For simple loaders, it is not typically necessary to create a special node subclass for the delegate, as you may provide an icon, cookies, and common actions without doing so.

System loaders

NetBeans installs a few extra "hidden" loaders into the loader pool, as are returned by DataLoaderPool.allLoaders(). Currently, these include among other things:

Cookies

Cookies provide a way for both data objects and nodes to indicate to NetBeans in a flexible fashion what types of (usually user-initiated) operations they are capable of supporting, and even to dynamically add and remove these capabilities.

What is a cookie?

A cookie is a design pattern used to separate the presentation of implementation of some interface from the actual object that implementation ought to be associated with. As such, it is a convenient way of removing the requirement that all interfaces presented by the cookie holder (either a data object or node) be actually implemented by the primary Java class of the object. If desired, the primary Java class can in fact be the cookie implementor, though this is a special case - some holder objects just declare themselves by default to hold all cookies their own class implements.

Take care not to confuse this NetBeans usage of the word with a different meaning sometimes used in computer science, that of an opaque callback object.

Originally there used to be a marker interface Node.Cookie, however it is no longer necessary to use it. Rather rely on getLookup method which can deal with plain Java objects. As such there are no requirements as to what makes a valid cookie - typically it will provide a small set of abstract operations, such as "opening", "compiling", "searching", etc.

Uses of cookies on nodes are not much different from uses on data objects; the initial cookie detection is done by Node.getLookup().lookup(...); however the default data node provides DataNode.getLookup(...) that looks inside MultiDataObject.getCookieSet().

Attaching and retrieving cookies

In short, there are a number of ways to attach cookies to either a node or data object (and you may listen for changes in the set of supported cookies, etc.). In all cases, the lookup(Class) method is used by the system to determine whether or not a given cookie is supported: thus, cookies are identified by their representation class, i.e. the Java class of the cookie interface. The system will expect to find a cookie object assignable to that representation class, and this object will have methods invoked on it to perform the proper action; it is the responsibility of that cookie object to associate itself to whatever node or data object is holding it.

Using cookie sets, it is possible to dynamically add and remove cookies from a holder (possible using AbstractLookup). This is appropriate for cookie types whose applicability may vary over time. For example, some objects may be compilable at certain times, but if they have been compiled recently and are already up-to-date, this support should be temporarily disabled. In the common case that an action (say, in the system Build menu for Compile) is sensitive to the cookie provided by the currently selected object (in this case the compilation cookie), this menu item can be automatically grayed out when necessary; and then reenabled when either the selection changes, or the object changes state so as to make the cookie appropriate once more, at which time the folder re-adds the cookie to its cookie set and LookupListener may observer the fired changes to this effect.

Cookie Usage Scenarios

There are various scenarios for ways in which you can use cookies, which demonstrate their flexibility. In the examples that follow, C means a cookie interface (or abstract class), O means a cookie holder (such as Node or DataObject), and S means a cookie support (concrete implementation). For example, signatures might look like this:

public interface C {
    void foo();
}
public class S implements C {
    private String param;
    public S(String param) {
        this.param = param;
    }
    public void foo() {
        System.out.println("foo: " + param);
    }
}
public class O extends DataObject {/* ... */}
Basic usage

Using cookies in the common way is pretty easy.

public class O extends DataObject {
    public O(FileObject fo) {
        // super...
        getCookieSet().assign(C1.class, new S1(fo));
        getCookieSet().assign(C2.class, new S2(this));
    }
}
public class S1 implements C1 {/* ... */}
public class S2 implements C2 {/* ... */}
// ...
DataObject o = DataObject.find(someFileObject);
// o instanceof O, in fact
C1 c1 = o.getLookup().lookup(C1.class);
// c1 instanceof S1
if (c1 != null) {
    c1.foo();
}
Multiple inheritance

Since cookies do not require language-level multiple inheritance, you can use subclassing naturally on supports.

public class S1 implements C1 {
    private String param;
    public S1(String param) {
        this.param = param;
    }
    public void foo1() {
        System.out.println("foo: " + transform(param));
    }
    /** Subclasses may customize. */
    protected String transform(String in) {
        return in;
    }
}
public abstract class S2 implements C2 {
    private String param;
    public S2(String param) {
        this.param = param;
    }
    public void foo2() {
        if (active()) {
            System.out.println("foo: " + param);
        }
    }
    /** Subclasses must implement. */
    protected abstract boolean active();
}
public class O {
    private int state;
    public O(String p) {
        state = INACTIVE; // initially
        getCookieSet().assign(C1.class, new MyS1(p));
        getCookieSet().assign(C2.class, new MyS2(p));
    }
    private static final class MyS1 extends S1 {
        public MyS1(String p) {super(p);}
        protected String transform(String in) {
            return in.toLowerCase();
        }
    }
    private final class MyS2 extends S2 {
        public MyS2(String p) {super(p);}
        protected boolean active() {
            return O.this.state == ACTIVE;
        }
    }
}
// ...
O o = new O("Hello");
C1 c1 = o.getLookup().lookup(C1.class);
if (c1 != null) c1.foo1();
// prints "foo: hello"
C2 c2 = o.getLookup().lookup(C2.class);
if (c2 != null) c2.foo2();
// does nothing: o is not yet active
Dynamically add or remove a cookie

Sometimes you want to change the capabilities of an object after its creation.

public class O {
    private boolean modified;
    public O(String param) {
        modified = false;
        // ...
    }
    private synchronized void markModified() {
        if (!modified) {
            // Newly modified, make it possible to save.
            // Note this will automatically fire a cookie change.
            getCookieSet().assign(SaveCookie.class, new SaveCookie() {
                public void save() {
                    doSave();
                }
            });
            modified = true;
        }
    }
    private synchronized void doSave() {
        if (modified) {
            // actually save...then:
            
            getCookieSet().assign(SaveCookie.class);
            modified = false;
        }
    }
}

Standard cookies and supports

Most standard cookies you would want to use exist in the org.openide.cookies package. Many of these have standard supports as well, frequently in org.openide.loaders. Using the Javadoc, the surest way to find cookies and supports is to look at Node.Cookie and browse the subinterfaces; for any cookie of interest, just look for implementing classes, which are usually supports. (A few cookie implementations are not of general utility, and so are not documented as being supports.) The standard supports generally take a file entry in their constructor, as they are designed for use by loaders attaching them to data objects.

Sometimes a cookie support that is applicable to multiple cookies may not implement any of them, leaving the choice of which to declare implementation of, to a subclass. This is the case with the abstract OpenSupport, which actually may be used for any or all of OpenCookie, ViewCookie, or CloseCookie, according to the needs of its subclass and holders. EditorSupport, for instance, uses only OpenCookie and CloseCookie.

Writing a cookie

Writing a new cookie requires no special knowledge - just extend create an interface and add its methods according to whatever you need done.

Often, in conjunction with writing a cookie you may want to create an action which is sensitive to that cookie. Then this action may be installed globally in the system (for example, in a menu bar, on a shortcut, etc.), and only activated when the current selection provides the cookie.

Writing a support for an existing cookie

Writing a support also does not require anything non-apparent. It should be a concrete class, its name conventionally ending in Support, implementing the cookie interface. Normally it should have one constructor, taking a MultiDataObject.Entry as principal argument, so as to encourage its use in the context of a loader; the file entry of course gives easy access to the file object it represents, as well as the data object via MultiDataObject.Entry.getDataObject().

If the support is designed to be usable from someone else's loader, and it is not obvious for which data objects using the support is possible, you may be well advised to include a public, static tester method in the support class indicating whether it would function correctly with a given file entry/data object. This way, an independently written loader could easily add the cookie with your support to any data object it had, without knowing the details of its prerequisites.

Using an existing support

Using an existing support is generally straightforward - assuming your data object is a subclass of MultiDataObject, you may just add a new instance of the support using MultiDataObject.setCookieSet(...) in the constructor, passing in the primary entry (most likely) from MultiDataObject.getPrimaryEntry().

Extended Clipboard

NetBeans implements an extended clipboard, which enhances the functions provided in the java.awt.datatransfer package. This implementation is to be found in org.openide.util.datatransfer.

Enabling complex customizations on both sides of a cut/copy-paste can be confusing; the Nodes API contains a detailed description of these operations as they pertain to nodes, which may be helpful.

Convertors

The extended clipboard, ExClipboard, provides the ability for a prebuilt transferable to support additional data flavors that it was not originally designed for; then the convertor supplies the implementation of these conversions.

To write a convertor, just implement ExClipboard.Convertor. The Javadoc should provide sufficient information on its sole method. Installing the convertor is easy; you can just add an instance of it to lookup.

Event notification

The extended clipboard supports Java Event-based notification of changes in the contents of the clipboard. Just register your listener with ExClipboard.addClipboardListener(...).

Multi-transfers

ExTransferable.Multi is a special transferable type used to transfer multiple objects (not necessarily of the same type) at once. It only supports one, virtual data flavor, ExTransferable.multiFlavor.

The data associated with multiFlavor in the ExTransferable.Multi will always be a special container object, MultiTransferObject. It is designed to permit access to its constituent real Transferables.

UML Diagrams

Data objects

DataObject UML

Data loaders

DataLoader UML