假设一个类定义了一个常量字段:
public class Foo {
public static final int CONSTANT_FIELD = 3;
}
假设注释接口声明如下:
public @interface Something {
int value();
}
最后,假设注释按如下方式使用:
@Something(Foo.CONSTANT_FIELD)
问题:在注释处理器中,我如何从设置@Something
的值中获得CONSTANT_FIELD
的元素?
编辑:在问题本身中包含一个具体的例子
我有一个注释是这样使用的:
@RuleDependency(recognizer = BQLParser.class,
rule = BQLParser.RULE_statement,
version = 0)
注释处理器需要知道RULE_statement
是在BQLParser
类中定义的常量。如果我可以通过设置注释的rule
属性直接访问BQLParser.RULE_statement
的Element
,那么就不需要recognizer
属性了。这个注释在实际应用程序中使用了数千次,recognizer
总是只是rule
常量的声明类型。解决这个问题可以将注释的使用简化为:
@RuleDependency(rule = BQLParser.RULE_statement, version = 0)
我可以使用Compiler Trees API实现这个特性。
-
更新pom.xml以包含以下配置文件,以确保tools.jar在默认未被引用的系统上被引用。
<profiles> <profile> <!-- Java 6 and earlier have java.vendor set to "Sun Microsystems Inc." --> <id>default-tools-6.jar</id> <activation> <property> <name>java.vendor</name> <value>Sun Microsystems Inc.</value> </property> </activation> <dependencies> <dependency> <groupId>com.sun</groupId> <artifactId>tools</artifactId> <version>1.6</version> <scope>system</scope> <systemPath>${java.home}/../lib/tools.jar</systemPath> </dependency> </dependencies> </profile> <profile> <!-- Java 7 and later have java.vendor set to "Oracle Corporation" --> <id>default-tools.jar</id> <activation> <property> <name>java.vendor</name> <value>Oracle Corporation</value> </property> </activation> <dependencies> <dependency> <groupId>com.sun</groupId> <artifactId>tools</artifactId> <version>1.6</version> <scope>system</scope> <systemPath>${java.home}/../lib/tools.jar</systemPath> </dependency> </dependencies> </profile> </profiles>
-
覆盖
Processor.init
以获得Trees
的实例。@Override public void init(ProcessingEnvironment processingEnv) { super.init(processingEnv); this.trees = Trees.instance(processingEnv); }
-
实现一个
TreePathScanner<TypeMirror, Void>
,用于获取在注释中分配给rule
属性的声明类型的TypeMirror
private class AnnotationVisitor extends TreePathScanner<TypeMirror, Void> { @Override public TypeMirror visitAnnotation(AnnotationTree node, Void p) { for (ExpressionTree expressionTree : node.getArguments()) { if (!(expressionTree instanceof AssignmentTree)) { continue; } AssignmentTree assignmentTree = (AssignmentTree)expressionTree; ExpressionTree variable = assignmentTree.getVariable(); if (!(variable instanceof IdentifierTree) || !((IdentifierTree)variable).getName().contentEquals("rule")) { continue; } return scan(expressionTree, p); } return null; } @Override public TypeMirror visitAssignment(AssignmentTree at, Void p) { return scan(at.getExpression(), p); } @Override public TypeMirror visitMemberSelect(MemberSelectTree mst, Void p) { return scan(mst.getExpression(), p); } @Override public TypeMirror visitIdentifier(IdentifierTree it, Void p) { return trees.getTypeMirror(this.getCurrentPath()); } }
-
提供
recognizer
属性的默认值。我希望这可能是null
,但Java明确禁止…/** * Gets the recognizer class where the dependent parser rules are defined. * This may reference the generated parser class directly, or for simplicity * in certain cases, any class derived from it. * <p> * If this value is not specified, the default value {@link Parser} * indicates that the declaring type of the constant value specified for * {@link #rule} should be used as the recognizer type. * </p> */ Class<? extends Recognizer<?, ?>> recognizer() default Parser.class;
-
更新收集有关应用于代码中特定
Element
实例的RuleDependency
注释的信息的代码,以首先尝试访问recognizer
属性,如果没有指定,则使用rule
属性中常量的声明类型代替。为简洁起见,在此代码示例中省略了错误处理。RuleDependency dependency = element.getAnnotation(RuleDependency.class); // first try to get the parser type from the annotation TypeMirror recognizerType = getRecognizerType(dependency); if (recognizerType != null && !recognizerType.toString().equals(Parser.class.getName())) { result.add(new Triple<RuleDependency, TypeMirror, Element>(dependency, recognizerType, element)); continue; } // fallback to compiler tree API AnnotationMirror annotationMirror = null; for (AnnotationMirror mirror : element.getAnnotationMirrors()) { if (processingEnv.getTypeUtils().isSameType(ruleDependencyTypeElement.asType(), mirror.getAnnotationType())) { annotationMirror = mirror; break; } } AnnotationValue annotationValue = null; for (Map.Entry<? extends ExecutableElement, ? extends AnnotationValue> entry : annotationMirror.getElementValues().entrySet()) { if (entry.getKey().getSimpleName().contentEquals("rule")) { annotationValue = entry.getValue(); break; } } TreePath treePath = trees.getPath(element, annotationMirror, annotationValue); AnnotationVisitor visitor = new AnnotationVisitor(); recognizerType = visitor.scan(treePath, null); result.add(new Triple<RuleDependency, TypeMirror, Element>(dependency, recognizerType, element));
如果rule
是一个int
,那么注释处理器就不可能找出在哪里定义了int
。但是,对于rule
字段,您可以使用enum
而不是int
,并在其中对规则进行分组,并引用它们的解析器。也许像这样:
解析器接口:
public interface RuleParser {
}
示例实现:
public class RuleParserImpl implements RuleParser {
}
规则枚举:
public enum Rule {
RULE_STATEMENT(RuleParserImpl.class);
private final Class<? extends RuleParser> ruleParserClass;
private Rule(Class<? extends RuleParser> ruleParser) {
this.ruleParserClass = ruleParser;
}
public Class<? extends RuleParser> getRuleParserClass() {
return ruleParserClass;
}
}
注释用enum代替int字段:
@Retention(RetentionPolicy.RUNTIME)
public @interface RuleDependency {
Rule rule();
}
使用例子:
@RuleDependency(rule = Rule.RULE_STATEMENT)
public class RuleProcessor {
public static void main(String[] args) {
RuleDependency ruleDependency = RuleProcessor.class.getAnnotation(RuleDependency.class);
Rule rule = ruleDependency.rule();
Class<? extends RuleParser> ruleParserClass = rule.getRuleParserClass();
System.out.println(ruleParserClass); //Prints "class RuleParserImpl"
}
}