JDT

Das Java Development Toolkit (JDT) stellt die Plugins für Eclipse bereit, die diese Plattform zu einer vollwertigen Java Entwicklungsumgebung machen.

JDT AST

Über JDT kann auf den abstrakten Syntaxbaum (AST) von Java Programmen zugegriffen werden. Damit kann der Source Code manipuliert werden und Code-Analysen durchgeführt werden.

Tutorials:

How does code look like in the JDT AST?

PACKAGE_DECLARATION is our own package. It consists of one or more QUALIFIED_NAME, all the names with at least one dot. A qualified name itself consists of simple names for each name between the dots. E.g. java.io.File has the qualified names java.io.File and java.io, then (in the AST) the simple names java, io, and File.

Same for IMPORT_DECLARATIONS. Imports with wildcards (java.io.*) have FQNs without the last dot and wildcard, i.e. java.io.

The class itself is a TYPE_DECLARATION. Then come the modifiers, then the class name as simple name.

Same for METHOD_DECLARATIONs.

Assignments are both, in that order, an EXPRESSION_STATEMENT and an ASSIGNMENT.

Instantiations with new() are *not* method calls, they are CLASS_INSTANCE_CREATIONs.

Invocations of methods with no class/package qualifiers (correct term?), i.e. foo() not bar.foo(), are a METHOD_INVOCATION followed immediately by *exactly one* simple name. These are "own" methods!

Invocations of methods on an object (foo.bar()) are METHOD_INVOCATIONs followed by two simple names. getFullyQualifiedName() is of no use here, it's only the methode name itself. One has to analyse the first simple name.

Even parameters in method invocations are simple names! So one.two.three() looks the same in the AST as one.two(three). You also cannot tell it from the traversal direction, it's "advance" in both cases. To recognize them, you can use the bindings.

The Javadoc preceding a method declaration is inside the METHOD_DECLARATION in the AST!

In method declarations, the parameters with their type, like the MyClass c in foo(MyClass c), are SINGLE_VARIABLE_DECLARATIONs.

Conditionals like IF_STATEMENT are the whole thing, including the body. the conditional itself - the whole one, if it has several parts - is an INFIX_EXPRESSION. Apparently, you cannot figure out what is actually tested except using string matching. Operators like && or > have no representation in the AST. (Not even as string or something, they only appear in the INFIX_EXPRESSION itself.)

For loops with iterators are a FOR_STATEMENT followed by a var. declaration (the Iterator it = foo.iterator())

INSTANCE_OF nodes have two leaves, the first for the variable as simple name, the second for the type as simple type.