Skip to main content

Notice: this Wiki will be going read only early in 2024 and edits will no longer be possible. Please see: https://gitlab.eclipse.org/eclipsefdn/helpdesk/-/wikis/Wiki-shutdown-plan for the plan.

Jump to: navigation, search

Difference between revisions of "JDT UI/Java8"

(TODOs)
(UI decisions)
(2 intermediate revisions by the same user not shown)
Line 151: Line 151:
  
 
*Use the following @since tag on all newly added members: "3.9 BETA_JAVA8"
 
*Use the following @since tag on all newly added members: "3.9 BETA_JAVA8"
 +
 +
= UI decisions =
 +
== TYPE_USE annotations ==
 +
TYPE_USE annotations are in a limbo state:
 +
*For the JLS8 and the JVM type system, they don't exist at all (they are just written to the class file, but they are never processed).
 +
*For tools like annotation processors or other compiler extensions that deal with TYPE_USE annotations, they are an important part of the type system. But since the JLS8 doesn't define the propagation of TYPE_USE annotations, there's no prescribed way to represent them e.g. in ITypeBindings for arbitrary expressions.
 +
 +
=> Conclusion: ITypeBindings don't carry TYPE_USE annotations.
 +
 +
Should TYPE_USE annotations show up in the Java model (IJavaElements)? The intention of the Java model is to provide a high-level structure for rendering elements in an outline and to serve as enclosing elements in search results. TYPE_USE annotations don't show up in type signatures and the compiler doesn't consult them. They are second-class citizens. The Java model cannot contain every detail of an element (e.g. it also doesn't contain comments). Furthermore, it's not clear how TYPE_USE annotations should be represented. They can't be part of a signature.
 +
 +
=> Conclusion: IJavaElements don't carry TYPE_USE annotations.
 +
 +
== Lambda expressions in the Java model ==
 +
Currently, anonymous classes are represented in the Java model. Should lambda expressions and method references also be part of the model?
 +
The problem with that is that lambda expressions are perceived more lightweight than anonymous classes, so lambda expressions will be used much more frequently. Method references are another form of implicit class declarations, since some forms already bind a receiver object.
 +
 +
Anonymous classes must be part of the Java model, because they can define multiple members (fields/methods) that should be visible in an Outline and that should support operations like "Open Super Implementation". OTOH, lambda expressions and method references don't have sub-members. However, a lambda expression can contain anonymous and local classes, and the lambda expression itself would be interesting as the enclosing element in a search result.
 +
 +
=> Conclusion: Lambda expressions are currently not represented in the Java model but might be added later.
  
 
= Things to remember/caveats  =
 
= Things to remember/caveats  =
Line 228: Line 248:
 
* Check helper methods in ASTNodes
 
* Check helper methods in ASTNodes
 
** e.g. getTopMostName(Name) is not correct any more to find an exception type
 
** e.g. getTopMostName(Name) is not correct any more to find an exception type
 +
 +
* Check all ASTVisitors that override one of the visit/endVisit methods for one of the changed AST node types
  
 
[[Category:JDT]]
 
[[Category:JDT]]

Revision as of 15:15, 27 March 2013

This page tracks the work in progress to add Java™ 8 support (mainly jsr308 "Type Annotations" and jsr335 "Lambda Expressions") into Eclipse JDT UI. See JDT Core/Java8 for the work in the JDT Core plug-ins.

Test matrix


Lambda Expressions Method references Default Methods Type Annotations / Receiver parameter
Typing, reconciler
Syntax coloring
Semantic highlighting
Mark occurrences
Outline
Content assist
Type hierarchy
Quick assists
Quick fixes
Clean ups
Refactoring operations (core)
Refactoring UIs
Formatter
Search
Javadoc (hover, view)
Source actions
Call hierarchy


Done
Tested and works, but can be improved (e.g. by adding Quick Assists or Quick Fixes)
Not done yet

(1) more work needed if spec changes

Work areas

  • set ASTProvider.SHARED_AST_LEVEL to JLS8 (bug 403927)
    • fix problems that need immediate handling
    • fix other references to JLS4
  • Annotations on types
    • Annotations that moved to another node in the AST (under discussion in bug 403834)
    • MethodDeclaration#thrownExceptions() -> thrownExceptionsTypes() (bug 403924)
    • Annotations on Extra Dimensions (bug 403923)
    • Receiver parameter on instance methods and inner class constructors (bug 404147)
    • Refactoring dialogs that allow to modify types
    • deferred: Render annotations in Javadoc hover/view. Do not render in Outline, etc. for now. (bug 403917)
  • Default Methods
  • Lambda Expressions
    • check code that walks the parent chain to find the enclosing Block or BodyDeclaration
  • Method References

IMPORTANT NOTE

  • The following lines must be added in all headers of modified files for Java™ 8 implementation:
 * This is an implementation of an early-draft specification developed under the Java
 * Community Process (JCP) and is made available for testing and evaluation purposes
 * only. The code is not compatible with any specification of the JCP.
 *

  • Use the following @since tag on all newly added members: "3.9 BETA_JAVA8"

UI decisions

TYPE_USE annotations

TYPE_USE annotations are in a limbo state:

  • For the JLS8 and the JVM type system, they don't exist at all (they are just written to the class file, but they are never processed).
  • For tools like annotation processors or other compiler extensions that deal with TYPE_USE annotations, they are an important part of the type system. But since the JLS8 doesn't define the propagation of TYPE_USE annotations, there's no prescribed way to represent them e.g. in ITypeBindings for arbitrary expressions.

=> Conclusion: ITypeBindings don't carry TYPE_USE annotations.

Should TYPE_USE annotations show up in the Java model (IJavaElements)? The intention of the Java model is to provide a high-level structure for rendering elements in an outline and to serve as enclosing elements in search results. TYPE_USE annotations don't show up in type signatures and the compiler doesn't consult them. They are second-class citizens. The Java model cannot contain every detail of an element (e.g. it also doesn't contain comments). Furthermore, it's not clear how TYPE_USE annotations should be represented. They can't be part of a signature.

=> Conclusion: IJavaElements don't carry TYPE_USE annotations.

Lambda expressions in the Java model

Currently, anonymous classes are represented in the Java model. Should lambda expressions and method references also be part of the model? The problem with that is that lambda expressions are perceived more lightweight than anonymous classes, so lambda expressions will be used much more frequently. Method references are another form of implicit class declarations, since some forms already bind a receiver object.

Anonymous classes must be part of the Java model, because they can define multiple members (fields/methods) that should be visible in an Outline and that should support operations like "Open Super Implementation". OTOH, lambda expressions and method references don't have sub-members. However, a lambda expression can contain anonymous and local classes, and the lambda expression itself would be interesting as the enclosing element in a search result.

=> Conclusion: Lambda expressions are currently not represented in the Java model but might be added later.

Things to remember/caveats

  • Goal of the first pass is to make Eclipse work with the new language features. For more advanced support (new quick fixes / refactorings / templates / ...), please file an enhancement request with the [1.8] tag.
  • Bugs that went into the BETA_JAVA8 branch should be RESOLVED/FIXED with Target Milestone "BETA_JAVA8" and the [1.8] tag in front of the summary.
  • For ASTNode changes:
    • Search for references to a changed AST Node and update code that relies on the concrete structure
    • Remember there are multiple ways to access AST node properties:
      • via getter/setter or List-valued accessor
      • via the StructuralPropertyDescriptor constant
      • for properties declared in an abstract class: via get*Property() instance method
    • Check references to superclasses of the changed/added node type. E.g. for IntersectionType, code that thinks it handles all known subtypes of Type needs to be adjusted.
    • Add new node types to switch or if-else-if statements
    • Think about cases where code could use StructuralPropertyDescriptors
  • Think about cases where new bindings can show up.

DOM AST node changes in JLS8

If everything is correct, then you should find these as references to ASTNode#unsupportedIn2_3_4() and #supportedOnlyIn2_3_4().

+AnnotatableType (abstract superclass):
  +annotations: List<Annotation> (also in subtypes PrimitiveType, ArrayType, SimpleType, QualifiedType, WildcardType)

+ExtraDimension
  +annotations: List<Annotation>

+LambdaExpression:
  +parentheses: boolean
  +parameters: List<SingleVariableDeclaration> or List<VariableDeclarationFragment>
  +body: Block or Expression
  +resolveMethodBinding(): IMethodBinding

~MethodDeclaration:
  -setExtraDimensions(int)
  +extraDimensions: List<ExtraDimension> (incl. annotations on extra dimensions)
  +receiverType: AnnotatableType
  +receiverQualifier: SimpleName
  -thrownExceptions: List<Name>
  +thrownExceptionTypes: List<Type>

~SingleVariableDeclaration:
  +varargsAnnotations: List<Annotation>

~TypeParameter:
  +annotations: List<Annotation>

~VariableDeclaration:
  +extraDimensions List<ExtraDimension> (also in subtypes SingleVariableDeclaration, VariableDeclarationFragment)

For JLS4, the changes were:

~TryStatement:
  +resources: List<VariableDeclarationExpression>

+UnionType
  +types: List<Type>

TODOs

  • 'default' flag:
    • show in the UI?
    • update JdtFlags?
  • The ITypeBinding of a TYPE_USE-annotated expression (e.g. ClassInstanceCreation or reference to a variable) doesn't contain the TYPE_USE annotations. References to "String s" and "@NonEmpty String ns" have identical type bindings. Should TYPE_USE annotations show up in reference bindings? Note that for some expressions, the annotated type is not defined by jsr 308, e.g. for the conditional expression here:
@Deprecated @NonEmpty String s = new Random().nextBoolean()
	? new @NonEmpty String("hi")
	: new String();
  • A VariableDeclarationFragment can now also show up in a LambdaExpression
    • check usages of VariableDeclarationFragment.getParent()
    • check usages of VariableDeclarationFragment that assume there's an associated type (e.g. ASTNodes.getType(VariableDeclaration))
  • Check helper methods in ASTNodes
    • e.g. getTopMostName(Name) is not correct any more to find an exception type
  • Check all ASTVisitors that override one of the visit/endVisit methods for one of the changed AST node types

Copyright © Eclipse Foundation, Inc. All Rights Reserved.