Eclipse Neon New and Noteworthy

A new Import Projects wizard is available via the File > Open Projects…​ menu and via the general Import…​ command under General > Projects from Folder or Archive.

Compared to other ways to import or create projects, this wizard aims at being a universal wizard that, from a source directory or archive, can detect and configure various kinds of projects.

This wizard relies on an extensible set of strategies that look at the content of the import source to detect file types, project layout patterns, or even look into some files in order to deduct which folders should be turned into projects and how those should be configured. Configuration usually involves setting up project natures and preferences.

The import strategies are contributed via the new extension point org.eclipse.ui.ide.projectConfigurator, so that other IDE components can contribute detection and configuration specific to the technology they’re supporting.

The Search > File... dialog has a new option to search in binary files as well.

By default, the option is disabled, and matches are only reported in files that have the Text content type, or in files that have no content type and don’t contain the null character (\u0000).

Auto-save of dirty editors is now available in Eclipse. The autosave option is disabled by default. A new autosave preference page (Preferences > General > Editors > Autosave) is available and allows to enable/disable the autosave and change the interval of autosave. The countdown is reset on keyboard activity, mouse click, or when a popup is displayed (e.g. content assist, preference page, …​).

Images now automatically scale on high-DPI monitors on Windows and Linux, similar to the Mac’s Retina support on OS X. In the absence of high-resolution images, available images are automatically scaled proportionately to the resolution of the monitor.

You can disable the CSS-based styling of the Eclipse IDE via Preferences > General > Appearance > Enable theming. This will prevent the Eclipse Workbench from rendering custom colors, shades, and borders, and may result in better performance.

The Full Screen feature is now also available on Windows and Linux. You can toggle the mode via shortcut (Alt+F11) or menu (Window > Appearance > Toggle Full Screen).

When Full Screen is activated, you’ll see a dialog which tells you how to turn it off again.

You can use the Toggle Visibility of all Toolbars command (via Quick Access: Ctrl+3 or Command+3) to hide all currently visible toolbars of the currently window. Selecting the command again, reveals these toolbars again. This allows the developer to maximize the space available for editors and views. If you minimize a stack after you selected this command, the minimized stack will be visible until you trigger the command to hide the toolbars again. This allows you to decide which minimized stacks are currently useful for you.

The following is a screenshot of the IDE with a maximized Java editor and several toolbars visible.

The next screenshot shows the same maximized editor but with hidden toolbars.

A Toggle Word Wrap button has been added to the workbench toolbar. Shortcut: Alt+Shift+Y.

By default, text editors are opened with word wrap disabled.

In text editors, you can now use Zoom In (Ctrl++ or Ctrl+=) and Zoom Out (Ctrl+-) commands to increase and decrease the font size.

Like a change in the General > Appearance > Colors and Fonts preference page, the commands persistently change the font size in all editors of the same type. If the editor type’s font is configured to use a default font, then that default font will be zoomed.

To temporarily zoom the editor font in text editors, use a "pinch" gesture on a touchpad. Put two fingers on the touchpad and move them apart or together.

To reset the original font size, rotate two fingers by at least 45°, or close and reopen the editor (Navigate > Back).

Those gestures only affect the current editor. Changes are neither propagated to other editors nor persisted.

The workspace selection dialog now allows you to start a previously selected workspace directly via a link. The path to the workspace is shortened. The full path is available if you hover over the link.

You can remove existing entries via the context menu.

On the Preferences > General > Editors > File Association page, you can now define an editor selection strategy for unassociated file types. Three strategies are provided out-of-the-box:

The implementation comes with a new extension-point org.eclipse.ui.ide.unknownEditorStrategy that allows external contributions of additional strategies.

Quick Access (Ctrl+3) is a small text field in the toolbar. You can use it to trigger any command in the Eclipse Workbench.

You can now restrict the search to Views, Commands, etc. by typing the category name followed by a colon. For example, to filter the list of all the views, start typing "Views: " in the search-box.

A few usability bugs have been fixed: The tooltip shows the keyboard shortcut, the number of search results per category is independent of the size of the proposals window, and the list with previous choices already opens when you click the field with the mouse.

The Welcome/Intro introduces a new Solstice theme that provides a new modern appearance and enhanced functionality.

In Eclipse 3.x, the Go to Workbench button on the Welcome page would minimize the Welcome page to the window status bar. This behavior has been restored.

The look of the toolbar in the MacOS default theme was adjusted to align with other MacOS applications.

To save space in the toolbar, the perspective switcher has been changed to show only the icons of perspectives by default. You can show the text again via the context menu.

The 'date' variable, supported in editor and code templates, now allows you to specify which format to use when inserting the current date (or time).

The 'word_selection' and 'line_selection' variables, supported in editor and code templates, now allow you to specify a default value to use if the current text selection is empty.

The workspace name is now shown at the beginning of the window title by default. The initial name is the name of the workspace directory.

When you export your preference settings via File > Export…​ > Preferences, the preferences are now sorted alphabetically in the resulting file.

Eclipse has been updated to include improved support for the Wayland display manager.

Launching Eclipse with recent Java 9 previews fails with NoClassDefFoundErrors for javax/annotation/PostConstruct and similar types.

The workaround is to add the VM arguments

either on the command line after -vmargs, or at the end of the eclipse.ini file on two separate lines.

The Java editor now offers default templates for creating "== null" and "!= null" checks.

You can now specify Info as a severity level for the problems detected by the Eclipse Java compiler.

A new decorator has been added to indicate information severity problems detected by the Eclipse Java Compiler.

You can now configure the severity of a compiler problem by invoking the new Quick Fix (Ctrl+1) which opens the Java > Compiler > Errors/Warnings preference page and highlights the configurable problem.

Content Assist now supports substring patterns. Enter any part of the desired proposal’s text, and Content Assist will find it! For example, completing on selection proposes all results containing selection as a substring.

This feature can be disabled using the Show substring matches option on the Java > Editor > Content Assist preference page.

A new option to remove redundant type arguments has been added under the Unnecessary Code group of the Clean Up profile.

In the new Parentheses tab, you can order the formatter to keep parentheses of various Java elements on separate lines, i.e. put a line break after the opening parenthesis and before the closing parenthesis. This can be done always, only when parentheses are not empty, or when their content is wrapped. There’s also an option to preserve existing positions, if you want to manually manage parentheses positions on a case-by-case basis.

In the Line Wrapping tab, you can set the wrapping policy for parameterized types, you can decide to wrap before or after operators in assignments and conditional expressions, and you can control the wrapping policy of 'for' loop headers.

The Align fields in columns feature in the Indentation section can now be configured to recognize groups separated by empty lines and align them independently.

In the New lines section, you can control if new lines should be added after annotations on enum constants.

You can now assign all parameters of a method or constructor to new fields at once using the new Quick Assist (Ctrl+1):

The Project Explorer view now groups referenced libraries in a new container node.

You can now hide empty library containers in the Project Explorer view by selecting the new filter in its Customize View…​ dialog.

You can now choose to add @Retention, @Target and @Documented meta-annotations along with their applicable values while creating a new Java annotation type.

You can now select multiple elements in views like Package Explorer and Outline and then search for References, Declarations, Implementors, and Read/Write Access (where applicable):

You can even select results in the Search view and start a new search for those elements. To follow method call chains, you’d probably use Open Call Hierarchy, though.

Assertions are now enabled by default for new JUnit launch configurations. You can disable this on the Preferences > Java > JUnit page.

Multiple versions of the same Java type can be available in a workspace, and each version can have breakpoints configured at different source locations. When debugging, JDT tries to determine the "right" set of breakpoints to install in the target VM. This analysis now uses Java project dependencies by default.

To always install all enabled breakpoints, you can disable the new option Preferences > Java > Debug > Do not install breakpoints from unrelated projects.

You can now configure annotation-based null analysis to use multiple sets of annotation types. This is relevant if your project uses 3rd party libraries that already use null annotations in their API, where the annotations used in a library may be different from those used in the current project.

Please open the dialog for configuring null annotation types from the project properties at Java Compiler > Errors/Warnings > Null analysis:

In the dialog that opens, only one primary set of annotations is supported; these are the annotations which JDT actively uses in Quick Fixes, error messages etc. These annotations must be present on the project’s build path.

You can now add an arbitrary number of secondary null annotation types, to be evaluated when reading class files external to the project. Within the configuration dialog, Content Assist is offered based on accessible annotation types, but for secondary annotation types unresolved names will be tolerated. This avoids the need to bloat the build path with transitive compile-time dependencies.

The interplay of null annotations and generics has been improved in several regards.

The basic concepts had already been documented in the online help, but not all that was documented was actually checked by null analysis. With the changes outlined below, null analysis based on type annotations is even more capable than previous versions.

(1) The first batch of contributed improvements implements what we call pessimistic analysis for free type variables, which affects implementors of generic types.

In order to allow clients to instantiate such a generic class with either a @NonNull or a @Nullable type as substitution for the "free type variable" <T>, the implementation of that class must account for the worst in both directions:

At the bottom of each problem hover, you will find a link to the corresponding configuration option, should you like to change the severity of this diagnostic. Note that configurability is mainly given for smooth migration from previous JDT version; conceptually, problems detected by pessimistic analysis should be treated as errors, because they invalidate the null-guarantees given by a generic type.

(2) The second improvement concerns consumption of generic types from a library, more specifically: invocations of library methods whose return type is given by a free type variable.

If the library type is instantiated with a @NonNull type as the type argument, we generally want to interpret the mentioned return type as non-null, too. This is problematic only for some "legacy" methods, which may return null without declaring so. The most prominent example is java.util.Map.get(K).

The analysis cannot see whether absence of a null annotation for such a return type is intentional (in the above sense of "free type variables"), or an omission that should be fixed. For that reason a new warning has been implemented to alert about this situation.

In the above example, both fields list and map provide @NonNull String as a type argument, hence the return types of both get methods are interpreted as non-null. For List this is desired, for Map it is a bug.

The dilemma can be resolved by adding a (possibly empty) external annotation file (.eea) to each affected library class. By using an empty annotation file, the user signals that all types in this class should be interpreted verbatim (like in the List case - use with care). In the other case, the missing @Nullable annotation should be added as an external annotation (like in the Map case).

In a project that is not yet configured to use external annotations for the library in question, the warning is demoted to level "information". Generally, the severity of this problem is configured using the option right below the one mentioned above, titled Unsafe '@NonNull' interpretation of free type variable from library.

(3) Finally, a small utility class, org.eclipse.jdt.annotation.Checks, has been included in the bundle org.eclipse.jdt.annotation_2.1.0 containing helper methods for typical idioms for improving and documenting null safety.

When a "potential null pointer access" problem is raised against a local variable, the reason is not always obvious. Perhaps null is assigned somewhere deep inside a complex control flow. If annotation-based null analysis is enabled, a new Quick Fix is offered (Ctrl+1), that adds a @NonNull annotation to the local variable.

While this may not produce correct code, it tells the compiler your intention to not allow null in this variable, and subsequently the compiler will answer with more helpful warnings or errors pointing to the root problem. In the given example, it will alert you of subtleties about using unconstrained type variables.

When code completion is invoked in a function call expression, appropriate substitutions are suggested for each of the function parameters.

In C files if "Type not found" error is detected, there is a new *Quick Fix* now to create a class based on that type, which opens the *New C Class Wizard*. Note: to enable quick user must enable "Type cannot be resolved" problem in "Code Analysis" settings

It is now possible to format either the whole file or only the edited lines when a source or a header file is saved.

It is now possible to suppress code analysis errors/warnings by adding comments in the code.

The comment is actually configurable using problem parameters (@suppress(<title>) is default).

For automatic remote debugging sessions, CDT now supports specifying extra options that will be passed to gdbserver when starting it. This way, users could activate debug messages (--debug), benefit of wrapper option (--wrapper) that gdbserver supports or use the other available command-line options of gdbserver.

The OS Resources view displays different information about the resources of the operating system, such as list of all processes running on the target or files opened by processes on target.

Now it is possible to attach to processes associated with selected resources directly from OS Resources view without resorting to "Connect to a process" button from Debug View.

This feature requires non-stop mode support to be activated.

OS Resources View lists a lot of text information. Now it is possible to copy selected lines to the clipboard to save for later comparison.

In an effort to reduce clutter, the Connect button was removed from the Debug toolbar. The action is still available from the Debug view context-menu as shown below. Furthermore, connecting to a process can now be done using the OS Resources view as described in this section.

To streamline the UI, the "New…​" button of the Connect dialogue which was used to create a new process has been removed. This action is available instead from the Debug view context-menu as shown below.

The 'Dockerfile Editor' provides users with content assist on the commands (ADD, COPY, RUN, etc.) as well as a customizable syntax highlighting.

Users can be given access to a shell environment within the container even if the container has a service running in the foreground. This is particularly useful for examining the container, and for debugging purposes.

Containers can be run with the --privileged flag, which gives them extended privileges, such as accessing all devices, and allowing them nearly all the same access to the host as regular processes running on the host (i.e., not in containers).

Sometimes the Docker instances that were configured in Eclipse are not running anymore (the Docker Machine is stopped, for example). In that case, the Docker Explorer view will display a different decorator for those connections, and an "Enable connection" button in the toolbar will let the user reconnect.

A new wizard lets the user edit the connection name, and settings. This is particularly useful when the connection belongs to a Docker Machine whose IP address changed upon restart.

Support now exists for performing actions upon Docker registries using authentication credentials. Users can simply set their credentials in the workspace preference page (Docker/Registry Accounts).

These credentials may be accessed through the Registry/Account combo box in the Push/Pull Wizards. The location of a registry can also be entered manually if it doesn’t require any credentials.

PHP 7 is the default version for new workspaces.

Final support for PHP 7 scalar type declarations.

The syntax highlighter now fully supports PHP 7 context sensitive keywords.

The PHP validator identifies non abstract methods without body.

Inappropriate use of multiple access modifiers (eg. public private functions)

The validator no longer analyzes PHP comments (except in PHPDoc comments).

Improved Code Assist for @param, @return, and @throws tags inside PHPDoc

The PHP indexer is significantly faster than in previous versions. The UI remains responsive when indexing is in progress.

Better task tags (e.g. @todo) detection

Methods and functions are visible as void only if they do not have return and yield statements. The otherwise have the declared return type or mixed.

Simple types (int, bool, string, …​) are included in the suggestions for for PHP >= 7 function return types.

More semantic highlighters are enabled by default, including class, deprecation, and constants.

Improved highlighting for array type.

Added support for instanceof and insteadof in class semantic highlighter.

Assign to local quick assist (ctrl/cmd + 2 + L)

Function / method return types are available in Content Assist proposals

Content Assist now supports colored labels

Buttons have been added to the editor toolbar to enable/disable Word wrap, Block selection, and Show white space characters.

Integration with the Project Explorer view has been improved, including dedicated support for PHP Project elements and specific actions, and support for traits when the "group by namespace" option is selected.

PHP Language Libraries have PHP Version badge

"Add PHP support…​" no longer visible in PHP-enabled projects (i.e. when it is unnecessary).

PHP Quick Fixes are available in Problems View.

Support for PHP exception breakpoints

Significant new work has been invested in the JavaScript development tools, starting with a new parser with EcmaScript 2015 (ES6) support. With this, the Eclipse IDE recognizes and support ES6 language constructs and provides better validation accuracy compared to the earlier parser.

Both Bower and npm are supported with creation wizards and launch shortcuts for executing update and install commands. Grunt and gulp build systems are accessible from within the Project Explorer view and can be executed via launch shortcuts.

Node.js runtimes are now directly supported (Eclipse Preferences > JavaScript > Runtimes) including a new launch configuration type can be used to run and debug Node.js applications. Node.js applications can be debugged directly from the editor via the Chromium V8 debugger, with full support for breakpoints and inspecting JavaScript variable values in the Variables view (and when you hover over an in-scope variable).

The new JSON editor provides support to edit some well-known JSON file types including package.json, bower.json, .bowerrc, and .jshintrc. This support includes the following features:

Acceleo introduced a new mechanism to launch code generators as part of a Maven build which is way easier and more robust by integrating with Tycho.

You will find several examples in this repository ranging from launchging an UML to C code generation to exporting a Modeling project as a website.

The Amalgam project brings a new view to navigate and explore models. Ecore models are supported right away and the view can be extended to support your domain specific models.

When using the Ecore graphical modeler references might now be displayed directly within the class shapes if it’s target is not in the diagram.

Inherited references will also be displayed if the super types are not in the diagram.

EMF Parsley is a lightweight framework that allows easy and quick UI development based upon EMF. EMF Parsley is built on top of the EMF Edit framework and it implements features like Trees, Forms and Table builders with standard JFace databinding, providing a complete component-based toolset. EMF Parsley can be configured to use all kinds of EMF persistence implementations (XMI, Teneo, CDO).

Moreover a DSL allows to easily customize several behaviors in each component.

The Neon release brings experimental support for comparing diagrams of Sirius based modelers.

With a right-click on a model element or resource in the viewer, a new context menu now offers you to merge all non-conflicting differences that are contained in the selected model element or resource with one click.

With local models (both writable):

With a read-only side (for example when comparing against a previous version in Git):

Each button now has a specific tooltip for each case depending on the selected element. It will display the best information depending on the context: which object will be affected by the merge action, and which side will be modified (or not) by the action.

With the new version of Sirius you can now define rich property views alongside the diagram definition without having to write a line of code.

This capability is introduced as an experimental feature for Eclipse Neon and will be extended in the next months based on the user community feedback.

Packs of SVG shapes can now be contributed through a plug-in to be reused in other modelers:

SVG rendering has been improved and is now pixel-perfect whatever the zoom level.

Before:

With Sirius 4.0:

It is now possible to define dash edge styles for both the nodes and the containers (like it was already possible for edges).

And the border node position relative to its parent can now be constrained on a specific side (NORTH, EAST, WEST or SOUTH).

The Link with editor support has been improved and is now bidirectional.

You now have the possibility to configure which elements to automatically select after a tool execution, using this it is possible to streamline the user experience by making sure that it is always possible to directly type right after the application of the tool.

The labels which are attached to an edge are now staying closer to it during a user modification.

A new mechanism is introduced to externalize the messages in a .odesign model.

The status of builds triggered by a patch set is shown on that patch set instead of appearing as comments. If the Hudson/Jenkins connector is installed, users can open the test results in the JUnit view directly from the patch set. This may not work if the Gerrit trigger plugin is configured with custom messages.

The task editor contains a section showing the reviews associated with the task, and indicates which reviews have unread incoming changes.

Inline comments are shown in threads, directly in the patch sets section in the review editor. This makes it easy to see the conversation taking place at a given line of code. Unread incoming comments are highlighed in blue.

Inline comments can now be edited and replied to inline, without opening a separate dialog.