[[semantics]] == Language Semantics [[semantics-intro]] === Introduction AspectJ extends Java by overlaying a concept of join points onto the existing Java semantics and adding a few new program elements to Java: A join point is a well-defined point in the execution of a program. These include method and constructor calls, field accesses and others described below. A pointcut picks out join points, and exposes some of the values in the execution context of those join points. There are several primitive pointcut designators, and others can be named and defined by the `pointcut` declaration. A piece of advice is code that executes at each join point in a pointcut. Advice has access to the values exposed by the pointcut. Advice is defined by `before`, `after`, and `around` declarations. Inter-type declarations form AspectJ's static crosscutting features, that is, is code that may change the type structure of a program, by adding to or extending interfaces and classes with new fields, constructors, or methods. Some inter-type declarations are defined through an extension of usual method, field, and constructor declarations, and other declarations are made with a new `declare` keyword. An aspect is a crosscutting type that encapsulates pointcuts, advice, and static crosscutting features. By type, we mean Java's notion: a modular unit of code, with a well-defined interface, about which it is possible to do reasoning at compile time. Aspects are defined by the `aspect` declaration. [[semantics-joinPoints]] === Join Points While aspects define types that crosscut, the AspectJ system does not allow completely arbitrary crosscutting. Rather, aspects define types that cut across principled points in a program's execution. These principled points are called join points. A join point is a well-defined point in the execution of a program. The join points defined by AspectJ are: Method call:: When a method is called, not including super calls of non-static methods. Method execution:: When the body of code for an actual method executes. Constructor call:: When an object is built and that object's initial constructor is called (i.e., not for `super` or `this` constructor calls). The object being constructed is returned at a constructor call join point, so its return type is considered to be the type of the object, and the object itself may be accessed with `after returning` advice. Constructor execution:: When the body of code for an actual constructor executes, after its this or super constructor call. The object being constructed is the currently executing object, and so may be accessed with the `this()` pointcut. The constructor execution join point for a constructor that calls a super constructor also includes any non-static initializers of enclosing class. No value is returned from a constructor execution join point, so its return type is considered to be `void`. Static initializer execution:: When the static initializer for a class executes. No value is returned from a static initializer execution join point, so its return type is considered to be `void`. Object pre-initialization:: Before the object initialization code for a particular class runs. This encompasses the time between the start of its first called constructor and the start of its parent's constructor. Thus, the execution of these join points encompass the join points of the evaluation of the arguments of `this()` and `super()` constructor calls. No value is returned from an object pre-initialization join point, so its return type is considered to be `void`. Object initialization:: When the object initialization code for a particular class runs. This encompasses the time between the return of its parent's constructor and the return of its first called constructor. It includes all the dynamic initializers and constructors used to create the object. The object being constructed is the currently executing object, and so may be accessed with the `this()` pointcut. No value is returned from a constructor execution join point, so its return type is considered to be `void`. Field reference:: When a non-constant field is referenced. [Note that references to constant fields (static final fields bound to a constant string object or primitive value) are not join points, since Java requires them to be inlined.] Field set:: When a field is assigned to. Field set join points are considered to have one argument, the value the field is being set to. No value is returned from a field set join point, so its return type is considered to be void. [Note that the initializations of constant fields (`static final` fields where the initializer is a constant string object or primitive value) are not join points, since Java requires their references to be inlined.] Handler execution:: When an exception handler executes. Handler execution join points are considered to have one argument, the exception being handled. No value is returned from a field set join point, so its return type is considered to be void. Advice execution:: When the body of code for a piece of advice executes. Each join point potentially has three pieces of state associated with it: the currently executing object, the target object, and an object array of arguments. These are exposed by the three state-exposing pointcuts, `this`, `target`, and `args`, respectively. Informally, the currently executing object is the object that a `this` expression would pick out at the join point. The target object is where control or attention is transferred to by the join point. The arguments are those values passed for that transfer of control or attention. [cols=",,,",options="header",] |=== |*Join Point* |*Current Object* |*Target Object* |*Arguments* |Method Call |executing object* |target object** |method arguments |Method Execution |executing object* |executing object* |method arguments |Constructor Call |executing object* |None |constructor arguments |Constructor Execution |executing object |executing object |constructor arguments |Static initializer execution |None |None |None |Object pre-initialization |None |None |constructor arguments |Object initialization |executing object |executing object |constructor arguments |Field reference |executing object* |target object** |None |Field assignment |executing object* |target object** |assigned value |Handler execution |executing object* |executing object* |caught exception |Advice execution |executing aspect |executing aspect |advice arguments |=== +++*+++ There is no executing object in static contexts such as static method bodies or static initializers. +++**+++ There is no target object for join points associated with static methods or fields. [[semantics-pointcuts]] === Pointcuts A pointcut is a program element that picks out join points and exposes data from the execution context of those join points. Pointcuts are used primarily by advice. They can be composed with boolean operators to build up other pointcuts. The primitive pointcuts and combinators provided by the language are: `call(MethodPattern)`:: Picks out each method call join point whose signature matches `_MethodPattern_`. `execution(MethodPattern)`:: Picks out each method execution join point whose signature matches `_MethodPattern_`. `get(FieldPattern)`:: Picks out each field reference join point whose signature matches `_FieldPattern_`. [Note that references to constant fields (static final fields bound to a constant string object or primitive value) are not join points, since Java requires them to be inlined.] `set(FieldPattern)`:: Picks out each field set join point whose signature matches `_FieldPattern_`. [Note that the initializations of constant fields (static final fields where the initializer is a constant string object or primitive value) are not join points, since Java requires their references to be inlined.] `call(ConstructorPattern)`:: Picks out each constructor call join point whose signature matches `_ConstructorPattern_`. `execution(ConstructorPattern)`:: Picks out each constructor execution join point whose signature matches `_ConstructorPattern_`. `initialization(ConstructorPattern)`:: Picks out each object initialization join point whose signature matches `_ConstructorPattern_`. `preinitialization(ConstructorPattern)`:: Picks out each object pre-initialization join point whose signature matches `_ConstructorPattern_`. `staticinitialization(TypePattern)`:: Picks out each static initializer execution join point whose signature matches `_TypePattern_`. `handler(TypePattern)`:: Picks out each exception handler join point whose signature matches `_TypePattern_`. `adviceexecution()`:: Picks out all advice execution join points. `within(TypePattern)`:: Picks out each join point where the executing code is defined in a type matched by `_TypePattern_`. `withincode(MethodPattern)`:: Picks out each join point where the executing code is defined in a method whose signature matches `_MethodPattern_`. `withincode(ConstructorPattern)`:: Picks out each join point where the executing code is defined in a constructor whose signature matches `_ConstructorPattern_`. `cflow(Pointcut)`:: Picks out each join point in the control flow of any join point `_P_` picked out by `_Pointcut_` , including `_P_` itself. `cflowbelow(Pointcut)`:: Picks out each join point in the control flow of any join point `_P_` picked out by `_Pointcut_`, but not `_P_` itself. `this(Type or Id)`:: Picks out each join point where the currently executing object (the object bound to `_this_`) is an instance of `_Type_` , or of the type of the identifier `_Id_` (which must be bound in the enclosing advice or pointcut definition). Will not match any join points from static contexts. `target(Type or Id)`:: Picks out each join point where the target object (the object on which a call or field operation is applied to) is an instance of `_Type_` , or of the type of the identifier `_Id_` (which must be bound in the enclosing advice or pointcut definition). Will not match any calls, gets, or sets of static members. `args(Type or Id, ...)`:: Picks out each join point where the arguments are instances of the appropriate type (or type of the identifier if using that form). A `_null_` argument is matched iff the static type of the argument (declared parameter type or field type) is the same as, or a subtype of, the specified args type. `PointcutId(TypePattern or Id, ...)`:: Picks out each join point that is picked out by the user-defined pointcut designator named by `_PointcutId_` . `if(BooleanExpression)`:: Picks out each join point where the boolean expression evaluates to `_true_` . The boolean expression used can only access static members, parameters exposed by the enclosing pointcut or advice, and `_thisJoinPoint_` forms. In particular, it cannot call non-static methods on the aspect or use return values or exceptions exposed by after advice. `! Pointcut`:: Picks out each join point that is not picked out by `_Pointcut_` . `Pointcut0 && Pointcut1`:: Picks out each join points that is picked out by both `_Pointcut0_` and `_Pointcut1_` . `Pointcut0 || Pointcut1`:: Picks out each join point that is picked out by either pointcuts. `_Pointcut0_` or `_Pointcut1_` . `( Pointcut )`:: Picks out each join points picked out by `_Pointcut_` . ==== Pointcut definition Pointcuts are defined and named by the programmer with the `pointcut` declaration. [source, java] .... pointcut publicIntCall(int i): call(public * *(int)) && args(i); .... A named pointcut may be defined in either a class or aspect, and is treated as a member of the class or aspect where it is found. As a member, it may have an access modifier such as `public` or `private`. [source, java] .... class C { pointcut publicCall(int i): call(public * *(int)) && args(i); } class D { pointcut myPublicCall(int i): C.publicCall(i) && within(SomeType); } .... Pointcuts that are not final may be declared abstract, and defined without a body. Abstract pointcuts may only be declared within abstract aspects. [source, java] .... abstract aspect A { abstract pointcut publicCall(int i); } .... In such a case, an extending aspect may override the abstract pointcut. [source, java] .... aspect B extends A { pointcut publicCall(int i): call(public Foo.m(int)) && args(i); } .... For completeness, a pointcut with a declaration may be declared `final`. Though named pointcut declarations appear somewhat like method declarations, and can be overridden in subaspects, they cannot be overloaded. It is an error for two pointcuts to be named with the same name in the same class or aspect declaration. The scope of a named pointcut is the enclosing class declaration. This is different than the scope of other members; the scope of other members is the enclosing class _body_. This means that the following code is legal: [source, java] .... aspect B percflow(publicCall()) { pointcut publicCall(): call(public Foo.m(int)); } .... ==== Context exposure Pointcuts have an interface; they expose some parts of the execution context of the join points they pick out. For example, the PublicIntCall above exposes the first argument from the receptions of all public unary integer methods. This context is exposed by providing typed formal parameters to named pointcuts and advice, like the formal parameters of a Java method. These formal parameters are bound by name matching. On the right-hand side of advice or pointcut declarations, in certain pointcut designators, a Java identifier is allowed in place of a type or collection of types. The pointcut designators that allow this are `this`, `target`, and `args`. In all such cases, using an identifier rather than a type does two things. First, it selects join points as based on the type of the formal parameter. So the pointcut [source, java] .... pointcut intArg(int i): args(i); .... picks out join points where an `int` (or a `byte`, `short`, or `char`; anything assignable to an `int`) is being passed as an argument. Second, though, it makes the value of that argument available to the enclosing advice or pointcut. Values can be exposed from named pointcuts as well, so [source, java] .... pointcut publicCall(int x): call(public *.*(int)) && intArg(x); pointcut intArg(int i): args(i); .... is a legal way to pick out all calls to public methods accepting an int argument, and exposing that argument. There is one special case for this kind of exposure. Exposing an argument of type Object will also match primitive typed arguments, and expose a "boxed" version of the primitive. So, [source, java] .... pointcut publicCall(): call(public *.*(..)) && args(Object); .... will pick out all unary methods that take, as their only argument, subtypes of Object (i.e., not primitive types like `int`), but [source, java] .... pointcut publicCall(Object o): call(public *.*(..)) && args(o); .... will pick out all unary methods that take any argument: And if the argument was an `int`, then the value passed to advice will be of type `java.lang.Integer`. The "boxing" of the primitive value is based on the _original_ primitive type. So in the following program [source, java] .... public class InstanceOf { public static void main(String[] args) { doInt(5); } static void doInt(int i) { } } aspect IntToLong { pointcut el(long l) : execution(* doInt(..)) && args(l); before(Object o) : el(o) { System.out.println(o.getClass()); } } .... The pointcut will match and expose the integer argument, but it will expose it as an `Integer`, not a `Long`. ==== Primitive pointcuts ===== Method-related pointcuts AspectJ provides two primitive pointcut designators designed to capture method call and execution join points. * `call( MethodPattern )` * `execution( MethodPattern )` ===== Field-related pointcuts AspectJ provides two primitive pointcut designators designed to capture field reference and set join points: * `get( FieldPattern )` * `set( FieldPattern )` All set join points are treated as having one argument, the value the field is being set to, so at a set join point, that value can be accessed with an `args` pointcut. So an aspect guarding a static integer variable x declared in type T might be written as [source, java] .... aspect GuardedX { static final int MAX_CHANGE = 100; before(int newval): set(static int T.x) && args(newval) { if (Math.abs(newval - T.x) > MAX_CHANGE) throw new RuntimeException(); } } .... ===== Object creation-related pointcuts AspectJ provides primitive pointcut designators designed to capture the initializer execution join points of objects. * `call( ConstructorPattern )` * `execution( ConstructorPattern )` * `initialization( ConstructorPattern )` * `preinitialization( ConstructorPattern )` ===== Class initialization-related pointcuts AspectJ provides one primitive pointcut designator to pick out static initializer execution join points. * `staticinitialization( TypePattern )` ===== Exception handler execution-related pointcuts AspectJ provides one primitive pointcut designator to capture execution of exception handlers: * `handler( TypePattern )` All handler join points are treated as having one argument, the value of the exception being handled. That value can be accessed with an `args` pointcut. So an aspect used to put `FooException` objects into some normal form before they are handled could be written as [source, java] .... aspect NormalizeFooException { before(FooException e): handler(FooException) && args(e) { e.normalize(); } } .... ===== Advice execution-related pointcuts AspectJ provides one primitive pointcut designator to capture execution of advice * `adviceexecution()` This can be used, for example, to filter out any join point in the control flow of advice from a particular aspect. [source, java] .... aspect TraceStuff { pointcut myAdvice(): adviceexecution() && within(TraceStuff); before(): call(* *(..)) && !cflow(myAdvice) { // do something } } .... ===== State-based pointcuts Many concerns cut across the dynamic times when an object of a particular type is executing, being operated on, or being passed around. AspectJ provides primitive pointcuts that capture join points at these times. These pointcuts use the dynamic types of their objects to pick out join points. They may also be used to expose the objects used for discrimination. * `this( Type or Id )` * `target( Type or Id )` The `this` pointcut picks out each join point where the currently executing object (the object bound to `this`) is an instance of a particular type. The `target` pointcut picks out each join point where the target object (the object on which a method is called or a field is accessed) is an instance of a particular type. Note that `target` should be understood to be the object the current join point is transfering control to. This means that the target object is the same as the current object at a method execution join point, for example, but may be different at a method call join point. * `args( Type or Id or "..", ...)` The args pointcut picks out each join point where the arguments are instances of some types. Each element in the comma-separated list is one of four things. If it is a type name, then the argument in that position must be an instance of that type. If it is an identifier, then that identifier must be bound in the enclosing advice or pointcut declaration, and so the argument in that position must be an instance of the type of the identifier (or of any type if the identifier is typed to Object). If it is the `*` wildcard, then any argument will match, and if it is the special wildcard `..`, then any number of arguments will match, just like in signature patterns. So the pointcut [source, java] .... args(int, .., String) .... will pick out all join points where the first argument is an `int` and the last is a `String`. ===== Control flow-based pointcuts Some concerns cut across the control flow of the program. The `cflow` and `cflowbelow` primitive pointcut designators capture join points based on control flow. * `cflow( Pointcut )` * `cflowbelow( Pointcut )` The `cflow` pointcut picks out all join points that occur between entry and exit of each join point `P` picked out by `Pointcut`, including `P` itself. Hence, it picks out the join points _in_ the control flow of the join points picked out by `Pointcut`. The `cflowbelow` pointcut picks out all join points that occur between entry and exit of each join point `P` picked out by `Pointcut`, but not including `P` itself. Hence, it picks out the join points _below_ the control flow of the join points picked out by `Pointcut`. ====== Context exposure from control flows The `cflow` and `cflowbelow` pointcuts may expose context state through enclosed `this`, `target`, and `args` pointcuts. Anytime such state is accessed, it is accessed through the _most recent_ control flow that matched. So the "current arg" that would be printed by the following program is zero, even though it is in many control flows. [source, java] .... class Test { public static void main(String[] args) { fact(5); } static int fact(int x) { if (x == 0) { System.err.println("bottoming out"); return 1; } else return x * fact(x - 1); } } aspect A { pointcut entry(int i): call(int fact(int)) && args(i); pointcut writing(): call(void println(String)) && ! within(A); before(int i): writing() && cflow(entry(i)) { System.err.println("Current arg is " + i); } } .... It is an error to expose such state through _negated_ control flow pointcuts, such as within `!cflowbelow(P)`. ===== Program text-based pointcuts While many concerns cut across the runtime structure of the program, some must deal with the lexical structure. AspectJ allows aspects to pick out join points based on where their associated code is defined. * `within( TypePattern )` * `withincode( MethodPattern )` * `withincode( ConstructorPattern )` The `within` pointcut picks out each join point where the code executing is defined in the declaration of one of the types in `TypePattern`. This includes the class initialization, object initialization, and method and constructor execution join points for the type, as well as any join points associated with the statements and expressions of the type. It also includes any join points that are associated with code in a type's nested types, and that type's default constructor, if there is one. The `withincode` pointcuts picks out each join point where the code executing is defined in the declaration of a particular method or constructor. This includes the method or constructor execution join point as well as any join points associated with the statements and expressions of the method or constructor. It also includes any join points that are associated with code in a method or constructor's local or anonymous types. ===== Expression-based pointcuts * `if( BooleanExpression )` The if pointcut picks out join points based on a dynamic property. its syntax takes an expression, which must evaluate to a boolean true or false. Within this expression, the `thisJoinPoint` object is available. So one (extremely inefficient) way of picking out all call join points would be to use the pointcut [source, java] .... if(thisJoinPoint.getKind().equals("call")) .... Note that the order of evaluation for pointcut expression components at a join point is undefined. Writing `if` pointcuts that have side-effects is considered bad style and may also lead to potentially confusing or even changing behavior with regard to when or if the test code will run. ==== Signatures One very important property of a join point is its signature, which is used by many of AspectJ's pointcut designators to select particular join points. ===== Methods Join points associated with methods typically have method signatures, consisting of a method name, parameter types, return type, the types of the declared (checked) exceptions, and some type that the method could be called on (below called the "qualifying type"). At a method call join point, the signature is a method signature whose qualifying type is the static type used to _access_ the method. This means that the signature for the join point created from the call `((Integer)i).toString()` is different than that for the call `((Object)i).toString()`, even if `i` is the same variable. At a method execution join point, the signature is a method signature whose qualifying type is the declaring type of the method. ===== Fields Join points associated with fields typically have field signatures, consisting of a field name and a field type. A field reference join point has such a signature, and no parameters. A field set join point has such a signature, but has a has a single parameter whose type is the same as the field type. ===== Constructors Join points associated with constructors typically have constructor signatures, consisting of a parameter types, the types of the declared (checked) exceptions, and the declaring type. At a constructor call join point, the signature is the constructor signature of the called constructor. At a constructor execution join point, the signature is the constructor signature of the currently executing constructor. At object initialization and pre-initialization join points, the signature is the constructor signature for the constructor that started this initialization: the first constructor entered during this type's initialization of this object. ===== Others At a handler execution join point, the signature is composed of the exception type that the handler handles. At an advice execution join point, the signature is composed of the aspect type, the parameter types of the advice, the return type (void for all but around advice) and the types of the declared (checked) exceptions. ==== Matching The `withincode`, `call`, `execution`, `get`, and `set` primitive pointcut designators all use signature patterns to determine the join points they describe. A signature pattern is an abstract description of one or more join-point signatures. Signature patterns are intended to match very closely the same kind of things one would write when declaring individual members and constructors. Method declarations in Java include method names, method parameters, return types, modifiers like static or private, and throws clauses, while constructor declarations omit the return type and replace the method name with the class name. The start of a particular method declaration, in class `Test`, for example, might be [source, java] .... class C { public final void foo() throws ArrayOutOfBoundsException { ... } } .... In AspectJ, method signature patterns have all these, but most elements can be replaced by wildcards. So [source, java] .... call(public final void C.foo() throws ArrayOutOfBoundsException) .... picks out call join points to that method, and the pointcut [source, java] .... call(public final void *.*() throws ArrayOutOfBoundsException) .... picks out all call join points to methods, regardless of their name name or which class they are defined on, so long as they take no arguments, return no value, are both `public` and `final`, and are declared to throw ``ArrayOutOfBoundsException``s. The defining type name, if not present, defaults to *, so another way of writing that pointcut would be [source, java] .... call(public final void *() throws ArrayOutOfBoundsException) .... The wildcard `..` indicates zero or more parameters, so [source, java] .... execution(void m(..)) .... picks out execution join points for void methods named `m`, of any number of arguments, while [source, java] .... execution(void m(.., int)) .... picks out execution join points for void methods named `m` whose last parameter is of type `int`. The modifiers also form part of the signature pattern. If an AspectJ signature pattern should match methods without a particular modifier, such as all non-public methods, the appropriate modifier should be negated with the `!` operator. So, [source, java] .... withincode(!public void foo()) .... picks out all join points associated with code in null non-public void methods named `foo`, while [source, java] .... withincode(void foo()) .... picks out all join points associated with code in null void methods named `foo`, regardless of access modifier. Method names may contain the * wildcard, indicating any number of characters in the method name. So [source, java] .... call(int *()) .... picks out all call join points to `int` methods regardless of name, but [source, java] .... call(int get*()) .... picks out all call join points to `int` methods where the method name starts with the characters "get". AspectJ uses the `new` keyword for constructor signature patterns rather than using a particular class name. So the execution join points of private null constructor of a class `C` defined to throw an `ArithmeticException` can be picked out with [source, java] .... execution(private C.new() throws ArithmeticException) .... ===== Matching based on the declaring type The signature-matching pointcuts all specify a declaring type, but the meaning varies slightly for each join point signature, in line with Java semantics. When matching for pointcuts `withincode`, `get`, and `set`, the declaring type is the class that contains the declaration. When matching method-call join points, the declaring type is the static type used to access the method. A common mistake is to specify a declaring type for the `call` pointcut that is a subtype of the originally-declaring type. For example, given the class [source, java] .... class Service implements Runnable { public void run() { ... } } .... the following pointcut [source, java] .... call(void Service.run()) .... would fail to pick out the join point for the code [source, java] .... ((Runnable) new Service()).run(); .... Specifying the originally-declaring type is correct, but would pick out any such call (here, calls to the `run()` method of any `Runnable`). In this situation, consider instead picking out the target type: [source, java] .... call(void run()) && target(Service) .... When matching method-execution join points, if the execution pointcut method signature specifies a declaring type, the pointcut will only match methods declared in that type, or methods that override methods declared in or inherited by that type. So the pointcut [source, java] .... execution(public void Middle.*()) .... picks out all method executions for public methods returning void and having no arguments that are either declared in, or inherited by, `Middle`, even if those methods are overridden in a subclass of `Middle`. So the pointcut would pick out the method-execution join point for `Sub.m()` in this code: [source, java] .... class Super { protected void m() { /*...*/ } } class Middle extends Super {} class Sub extends Middle { public void m() { /*...*/ } } .... ===== Matching based on the `throws` clause Type patterns may be used to pick out methods and constructors based on their `throws` clauses. This allows the following two kinds of extremely wildcarded pointcuts: [source, java] .... pointcut throwsMathlike(): // each call to a method with a throws clause containing at least // one exception exception with "Math" in its name. call(* *(..) throws *..*Math*); pointcut doesNotThrowMathlike(): // each call to a method with a throws clause containing no // exceptions with "Math" in its name. call(* *(..) throws !*..*Math*); .... A `ThrowsClausePattern` is a comma-separated list of ``ThrowsClausePatternItem``s, where [source, text] .... ThrowsClausePatternItem := [ ! ] TypeNamePattern .... A `ThrowsClausePattern` matches the `throws` clause of any code member signature. To match, each `ThrowsClausePatternItem` must match the `throws` clause of the member in question. If any item doesn't match, then the whole pattern doesn't match. If a `ThrowsClausePatternItem` begins with `!`, then it matches a particular `throws` clause if and only if _none_ of the types named in the `throws` clause is matched by the `TypeNamePattern`. If a `ThrowsClausePatternItem` does not begin with `!`, then it matches a throws clause if and only if _any_ of the types named in the `throws` clause is matched by the `TypeNamePattern`. The rule for `!` matching has one potentially surprising property, in that these two pointcuts . `call(* *(..) throws !IOException)` . `call(* *(..) throws (!IOException))` will match differently on calls to [source, java] .... void m() throws RuntimeException, IOException {} .... [1] will *not* match the method `m()`, because ``m``'s throws clause declares that it `throws IOException`. [2] *will* match the method `m()`, because ``m``'s throws clause declares that it throws some exception which does not match `IOException`, i.e. `RuntimeException`. ==== Type patterns Type patterns are a way to pick out collections of types and use them in places where you would otherwise use only one type. The rules for using type patterns are simple. ===== Exact type pattern First, all type names are also type patterns. So `Object`, `java.util.HashMap`, `Map.Entry`, `int` are all type patterns. If a type pattern is an exact type - if it doesn't include a wildcard - then the matching works just like normal type lookup in Java: * Patterns that have the same names as primitive types (like `int`) match those primitive types. * Patterns that are qualified by package names (like `java.util.HashMap`) match types in other packages. * Patterns that are not qualified (like `HashMap`) match types that are resolved by Java's normal scope rules. So, for example, `HashMap` might match a package-level type in the same package or a type that have been imported with Java's `import` form. But it would not match `java.util.HashMap` unless the aspect were in `java.util` or the type had been imported. So exact type patterns match based on usual Java scope rules. ===== Type name patterns There is a special type name, `\*`, which is also a type pattern. `*` picks out all types, including primitive types. So [source, java] .... call(void foo(*)) .... picks out all call join points to void methods named foo, taking one argument of any type. Type names that contain the two wildcards `\*` and `..` are also type patterns. The `*` wildcard matches zero or more characters characters except for `.`, so it can be used when types have a certain naming convention. So [source, java] .... handler(java.util.*Map) .... picks out the types `java.util.Map` and `java.util.java.util.HashMap`, among others, and [source, java] .... handler(java.util.*) .... picks out all types that start with `java.util.` and don't have any more ``.``s, that is, the types in the `java.util` package, but not inner types (such as `java.util.Map.Entry`). The `..` wildcard matches any sequence of characters that start and end with a `.`, so it can be used to pick out all types in any subpackage, or all inner types. So [source, java] .... within(com.xerox..*) .... picks out all join points where the code is in any declaration of a type whose name begins with `com.xerox.`. Type patterns with wildcards do not depend on Java's usual scope rules - they match against all types available to the weaver, not just those that are imported into an Aspect's declaring file. ===== Subtype patterns It is possible to pick out all subtypes of a type (or a collection of types) with the `+` wildcard. The `+` wildcard follows immediately a type name pattern. So, while [source, java] .... call(Foo.new()) .... picks out all constructor call join points where an instance of exactly type `Foo` is constructed, [source, java] .... call(Foo+.new()) .... picks out all constructor call join points where an instance of any subtype of `Foo` (including `Foo` itself) is constructed, and the unlikely [source, java] .... call(*Handler+.new()) .... picks out all constructor call join points where an instance of any subtype of any type whose name ends in `Handler` is constructed. ===== Array type patterns A type name pattern or subtype pattern can be followed by one or more sets of square brackets to make array type patterns. So `Object[]` is an array type pattern, and so is `com.xerox..*[][]`, and so is `Object+[]`. ===== Type patterns Type patterns are built up out of type name patterns, subtype patterns, and array type patterns, and constructed with boolean operators `&&`, `||`, and `!`. So [source, java] .... staticinitialization(Foo || Bar) .... picks out the static initializer execution join points of either `Foo` or `Bar`, and [source, java] .... call((Foo+ && ! Foo).new(..)) .... picks out the constructor call join points when a subtype of `Foo`, but not `Foo` itself, is constructed. ==== Pattern Summary Here is a summary of the pattern syntax used in AspectJ: [source, text] .... MethodPattern = [ModifiersPattern] TypePattern [TypePattern . ] IdPattern (TypePattern | ".." , ... ) [ throws ThrowsPattern ] ConstructorPattern = [ModifiersPattern ] [TypePattern . ] new (TypePattern | ".." , ...) [ throws ThrowsPattern ] FieldPattern = [ModifiersPattern] TypePattern [TypePattern . ] IdPattern ThrowsPattern = [ ! ] TypePattern , ... TypePattern = IdPattern [ + ] [ [] ... ] | ! TypePattern | TypePattern && TypePattern | TypePattern || TypePattern | ( TypePattern ) IdPattern = Sequence of characters, possibly with special * and .. wildcards ModifiersPattern = [ ! ] JavaModifier ... .... [[semantics-advice]] === Advice Each piece of advice is of the form [source, text] .... [ strictfp ] AdviceSpec [ throws TypeList ] : Pointcut { Body } .... where `AdviceSpec` is one of * `before( Formals )` * `after( Formals ) returning [ ( Formal ) ]` * `after( Formals ) throwing [ ( Formal ) ]` * `after( Formals )` * `Type around( Formals )` and where `Formal` refers to a variable binding like those used for method parameters, of the form `Type` `Variable-Name`, and `Formals` refers to a comma-delimited list of `Formal`. Advice defines crosscutting behavior. It is defined in terms of pointcuts. The code of a piece of advice runs at every join point picked out by its pointcut. Exactly how the code runs depends on the kind of advice. AspectJ supports three kinds of advice. The kind of advice determines how it interacts with the join points it is defined over. Thus AspectJ divides advice into that which runs *before* its join points, that which runs *after* its join points, and that which runs *in place of (or "around")* its join points. While `before` advice is relatively unproblematic, there can be three interpretations of `after` advice: After the execution of a join point completes normally, after it throws an exception, or after it does either one. AspectJ allows `after` advice for any of these situations: [source, java] .... aspect A { pointcut publicCall(): call(public Object *(..)); after() returning (Object o): publicCall() { System.out.println("Returned normally with " + o); } after() throwing (Exception e): publicCall() { System.out.println("Threw an exception: " + e); } after(): publicCall(){ System.out.println("Returned or threw an Exception"); } } .... `after returning` advice may not care about its returned object, in which case it may be written [source, java] .... after() returning: call(public Object *(..)) { System.out.println("Returned normally"); } .... If `after returning` does expose its returned object, then the type of the parameter is considered to be an `instanceof`-like constraint on the advice: it will run only when the return value is of the appropriate type. A value is of the appropriate type if it would be assignable to a variable of that type, in the Java sense. That is, a `byte` value is assignable to a `short` parameter but not vice-versa, an `int` is assignable to a `float` parameter, `boolean` values are only assignable to `boolean` parameters, and reference types work by `instanceof`. There are two special cases: If the exposed value is typed to `Object`, then the advice is not constrained by that type: the actual return value is converted to an object type for the body of the advice: `int` values are represented as `java.lang.Integer` objects, etc, and no value (from `void` methods, for example) is represented as `null`. Secondly, the `null` value is assignable to a parameter `T` if the join point _could_ return something of type `T`. `around` advice runs in place of the join point it operates over, rather than before or after it. Because `around` is allowed to return a value, it must be declared with a return type, like a method. Thus, a simple use of `around` advice is to make a particular method constant: [source, java] .... aspect A { int around(): call(int C.foo()) { return 3; } } .... Within the body of `around` advice, though, the computation of the original join point can be executed with the special syntax [source, java] .... proceed( ... ) .... The `proceed` form takes as arguments the context exposed by the around's pointcut, and returns whatever the around is declared to return. So the following around advice will double the second argument to `foo` whenever it is called, and then halve its result: [source, java] .... aspect A { int around(int i): call(int C.foo(Object, int)) && args(i) { int newi = proceed(i*2) return newi/2; } } .... If the return value of `around` advice is typed to `Object`, then the result of proceed is converted to an object representation, even if it is originally a primitive value. And when the advice returns an `Object` value, that value is converted back to whatever representation it was originally. So another way to write the doubling and halving advice is: [source, java] .... aspect A { Object around(int i): call(int C.foo(Object, int)) && args(i) { Integer newi = (Integer) proceed(i*2) return new Integer(newi.intValue() / 2); } } .... Any occurence of `proceed(..)` within the body of around advice is treated as the special `proceed` form (even if the aspect defines a method named `proceed`), unless a target other than the aspect instance is specified as the recipient of the call. For example, in the following program the first call to `proceed` will be treated as a method call to the `ICanProceed` instance, whereas the second call to `proceed` is treated as the special `proceed` form. [source, java] .... aspect A { Object around(ICanProceed canProceed) : execution(* *(..)) && this(canProceed) { canProceed.proceed(); // a method call return proceed(canProceed); // the special proceed form } private Object proceed(ICanProceed canProceed) { // this method cannot be called from inside the body of around advice // in the aspect } } .... In all kinds of advice, the parameters of the advice behave exactly like method parameters. In particular, assigning to any parameter affects only the value of the parameter, not the value that it came from. This means that [source, java] .... aspect A { after() returning (int i): call(int C.foo()) { i = i * 2; } } .... will _not_ double the returned value of the advice. Rather, it will double the local parameter. Changing the values of parameters or return values of join points can be done by using `around` advice. With `proceed(..)` it is possible to change the values used by less-precedent advice and the underlying join point by supplying different values for the variables. For example, this aspect replaces the string bound to `s` in the named pointcut `privateData`: [source, java] .... aspect A { Object around(String s): MyPointcuts.privateData(s) { return proceed("private data"); } } .... If you replace an argument to `proceed(..)`, you can cause a `ClassCastException` at runtime when the argument refers to a supertype of the actual type and you do not supply a reference of the actual type. In the following aspect, the around advice replaces the declared target `List` with an `ArrayList`. This is valid code at compile-time since the types match. [source, java] .... import java.util.*; aspect A { Object around(List list): call(* List+.*()) && target(list) { return proceed(new ArrayList()); } } .... But imagine a simple program where the actual target is `LinkedList`. In this case, the advice would cause a `ClassCastException` at runtime, and `peek()` is not declared in `ArrayList`. [source, java] .... public class Test { public static void main(String[] args) { new LinkedList().peek(); } } .... The `ClassCastException` can occur even in situations where it appears to be unnecessary, e.g., if the program is changed to call `size()`, declared in `List`: [source, java] .... public class Test { public static void main(String[] args) { new LinkedList().size(); } } .... There will still be a `ClassCastException` because it is impossible to prove that there won't be a runtime binary-compatible change in the hierarchy of `LinkedList` or some other advice on the join point that requires a `LinkedList`. ==== Advice modifiers The `strictfp` modifier is the only modifier allowed on advice, and it has the effect of making all floating-point expressions within the advice be FP-strict. ==== Advice and checked exceptions An advice declaration must include a `throws` clause listing the checked exceptions the body may throw. This list of checked exceptions must be compatible with each target join point of the advice, or an error is signalled by the compiler. For example, in the following declarations: [source, java] .... import java.io.FileNotFoundException; class C { int i; int getI() { return i; } } aspect A { before(): get(int C.i) { throw new FileNotFoundException(); } before() throws FileNotFoundException: get(int C.i) { throw new FileNotFoundException(); } } .... both pieces of advice are illegal. The first because the body throws an undeclared checked exception, and the second because field get join points cannot throw ``FileNotFoundException``s. The exceptions that each kind of join point in AspectJ may throw are: method call and execution:: the checked exceptions declared by the target method's `throws` clause. constructor call and execution:: the checked exceptions declared by the target constructor's `throws` clause. field get and set:: no checked exceptions can be thrown from these join points. exception handler execution:: the exceptions that can be thrown by the target exception handler. static initializer execution:: no checked exceptions can be thrown from these join points. pre-initialization and initialization:: any exception that is in the `throws` clause of all constructors of the initialized class. advice execution:: any exception that is in the `throws` clause of the advice. ==== Advice precedence Multiple pieces of advice may apply to the same join point. In such cases, the resolution order of the advice is based on advice precedence. ===== Determining precedence There are a number of rules that determine whether a particular piece of advice has precedence over another when they advise the same join point. If the two pieces of advice are defined in different aspects, then there are three cases: * If aspect `A` is matched earlier than aspect `B` in some `declare precedence` form, then all advice in concrete aspect `A` has precedence over all advice in concrete aspect `B` when they are on the same join point. * Otherwise, if aspect `A` is a subaspect of aspect `B`, then all advice defined in `A` has precedence over all advice defined in `B`. So, unless otherwise specified with `declare precedence`, advice in a subaspect has precedence over advice in a superaspect. * Otherwise, if two pieces of advice are defined in two different aspects, it is undefined which one has precedence. If the two pieces of advice are defined in the same aspect, then there are two cases: * If either are `after` advice, then the one that appears later in the aspect has precedence over the one that appears earlier. * Otherwise, then the one that appears earlier in the aspect has precedence over the one that appears later. These rules can lead to circularity, such as [source, java] .... aspect A { before(): execution(void main(String[] args)) {} after(): execution(void main(String[] args)) {} before(): execution(void main(String[] args)) {} } .... such circularities will result in errors signalled by the compiler. ===== Effects of precedence At a particular join point, advice is ordered by precedence. A piece of `around` advice controls whether advice of lower precedence will run by calling `proceed`. The call to `proceed` will run the advice with next precedence, or the computation under the join point if there is no further advice. A piece of `before` advice can prevent advice of lower precedence from running by throwing an exception. If it returns normally, however, then the advice of the next precedence, or the computation under the join pint if there is no further advice, will run. Running `after returning` advice will run the advice of next precedence, or the computation under the join point if there is no further advice. Then, if that computation returned normally, the body of the advice will run. Running `after throwing` advice will run the advice of next precedence, or the computation under the join point if there is no further advice. Then, if that computation threw an exception of an appropriate type, the body of the advice will run. Running `after` advice will run the advice of next precedence, or the computation under the join point if there is no further advice. Then the body of the advice will run. ==== Reflective access to the join point Three special variables are visible within bodies of advice and within `if()` pointcut expressions: `thisJoinPoint`, `thisJoinPointStaticPart`, and `thisEnclosingJoinPointStaticPart`. Each is bound to an object that encapsulates some of the context of the advice's current or enclosing join point. These variables exist because some pointcuts may pick out very large collections of join points. For example, the pointcut [source, java] .... pointcut publicCall(): call(public * *(..)); .... picks out calls to many methods. Yet the body of advice over this pointcut may wish to have access to the method name or parameters of a particular join point. * `thisJoinPoint` is bound to a complete join point object. * `thisJoinPointStaticPart` is bound to a part of the join point object that includes less information, but for which no memory allocation is required on each execution of the advice. It is equivalent to `thisJoinPoint.getStaticPart()`. * `thisEnclosingJoinPointStaticPart` is bound to the static part of the join point enclosing the current join point. Only the static part of this enclosing join point is available through this mechanism. Standard Java reflection uses objects from the `java.lang.reflect` hierarchy to build up its reflective objects. Similarly, AspectJ join point objects have types in a type hierarchy. The type of objects bound to `thisJoinPoint` is `org.aspectj.lang.JoinPoint`, while `thisStaticJoinPoint` is bound to objects of interface type `org.aspectj.lang.JoinPoint.StaticPart`. [[semantics-declare]] === Static crosscutting Advice declarations change the behavior of classes they crosscut, but do not change their static type structure. For crosscutting concerns that do operate over the static structure of type hierarchies, AspectJ provides inter-type member declarations and other `declare` forms. ==== Inter-type member declarations AspectJ allows the declaration of members by aspects that are associated with other types. An inter-type method declaration looks like * `[ Modifiers ] Type OnType . Id ( Formals ) [ ThrowsClause ] { Body }` * `abstract [ Modifiers ] Type OnType . Id ( Formals ) [ ThrowsClause ] ;` The effect of such a declaration is to make `OnType` support the new method. Even if `OnType` is an interface. Even if the method is neither public nor abstract. So the following is legal AspectJ code: [source, java] .... interface Iface {} aspect A { private void Iface.m() { System.err.println("I'm a private method on an interface"); } void worksOnI(Iface iface) { // calling a private method on an interface iface.m(); } } .... An inter-type constructor declaration looks like * `[ Modifiers ] OnType . new ( Formals ) [ ThrowsClause ] { Body }` The effect of such a declaration is to make `OnType` support the new constructor. It is an error for `OnType` to be an interface. Inter-type declared constructors cannot be used to assign a value to a final variable declared in `OnType`. This limitation significantly increases the ability to both understand and compile the `OnType` class and the declaring aspect separately. Note that in the Java language, classes that define no constructors have an implicit no-argument constructor that just calls `super()`. This means that attempting to declare a no-argument inter-type constructor on such a class may result in a conflict, even though it _looks_ like no constructor is defined. An inter-type field declaration looks like one of * `[ Modifiers ] Type OnType . Id = Expression ;` * `[ Modifiers ] Type OnType . Id ;` The effect of such a declaration is to make `OnType` support the new field. Even if `OnType` is an interface. Even if the field is neither public, nor static, nor final. The initializer, if any, of an inter-type field declaration runs before the class-local initializers defined in its target class. Any occurrence of the identifier `this` in the body of an inter-type constructor or method declaration, or in the initializer of an inter-type field declaration, refers to the `OnType` object rather than to the aspect type; it is an error to access `this` in such a position from a `static` inter-type member declaration. ==== Access modifiers Inter-type member declarations may be `public` or `private`, or have default (package-protected) visibility. AspectJ does not provide protected inter-type members. The access modifier applies in relation to the aspect, not in relation to the target type. So a private inter-type member is visible only from code that is defined within the declaring aspect. A default-visibility inter-type member is visible only from code that is defined within the declaring aspect's package. Note that a declaring a private inter-type method (which AspectJ supports) is very different from inserting a private method declaration into another class. The former allows access only from the declaring aspect, while the latter would allow access only from the target type. Java serialization, for example, uses the presense of a private method `void writeObject(ObjectOutputStream)` for the implementation of `java.io.Serializable`. A private inter-type declaration of that method would not fulfill this requirement, since it would be private to the aspect, not private to the target type. The access modifier of abstract inter-type methods has one constraint: It is illegal to declare an abstract non-public inter-type method on a public interface. This is illegal because it would say that a public interface has a constraint that only non-public implementors must fulfill. This would not be compatible with Java's type system. ==== Conflicts Inter-type declarations raise the possibility of conflicts among locally declared members and inter-type members. For example, assuming `otherPackage` is not the package containing the aspect `A`, the code [source, java] .... aspect A { private Registry otherPackage.onType.r; public void otherPackage.onType.register(Registry r) { r.register(this); this.r = r; } } .... declares that `onType` in `otherPackage` has a field `r`. This field, however, is only accessible from the code inside of aspect `A`. The aspect also declares that `onType` has a method "`register`", but makes this method accessible from everywhere. If `onType` already defines a private or package-protected field `r`, there is no conflict: The aspect cannot see such a field, and no code in `otherPackage` can see the inter-type `r`. If `onType` defines a public field `r`, there is a conflict: The expression [source, java] .... this.r = r .... is an error, since it is ambiguous whether the private inter-type `r` or the public locally-defined `r` should be used. If `onType` defines a method `register(Registry)` there is a conflict, since it would be ambiguous to any code that could see such a defined method which `register(Registry)` method was applicable. Conflicts are resolved as much as possible as per Java's conflict resolution rules: * A subclass can inherit multiple fields from its superclasses, all with the same name and type. However, it is an error to have an ambiguous reference to a field. * A subclass can only inherit multiple methods with the same name and argument types from its superclasses if only zero or one of them is concrete (i.e., all but one is abstract, or all are abstract). Given a potential conflict between inter-type member declarations in different aspects, if one aspect has precedence over the other its declaration will take effect without any conflict notice from compiler. This is true both when the precedence is declared explicitly with `declare precedence` as well as when when sub-aspects implicitly have precedence over their super-aspect. ==== Extension and Implementation An aspect may change the inheritance hierarchy of a system by changing the superclass of a type or adding a superinterface onto a type, with the `declare parents` form. * `declare parents: TypePattern extends Type ;` * `declare parents: TypePattern implements TypeList ;` For example, if an aspect wished to make a particular class runnable, it might define appropriate inter-type `void run()` method, but it should also declare that the class fulfills the `Runnable` interface. In order to implement the methods in the `Runnable` interface, the inter-type `run()` method must be public: [source, java] .... aspect A { declare parents: SomeClass implements Runnable; public void SomeClass.run() { ... } } .... ==== Interfaces with members Through the use of inter-type members, interfaces may now carry (non-public-static-final) fields and (non-public-abstract) methods that classes can inherit. Conflicts may occur from ambiguously inheriting members from a superclass and multiple superinterfaces. Because interfaces may carry non-static initializers, each interface behaves as if it has a zero-argument constructor containing its initializers. The order of super-interface instantiation is observable. We fix this order with the following properties: A supertype is initialized before a subtype, initialized code runs only once, and the initializers for a type's superclass are run before the initializers for its superinterfaces. Consider the following hierarchy where {`Object`, `C`, `D`, `E`} are classes, {`M`, `N`, `O`, `P`, `Q`} are interfaces. [source, text] .... Object M O \ / \ / C N Q \ / / D P \ / E .... when a new `E` is instantiated, the initializers run in this order: [source, text] .... Object M C O N D Q P E .... ==== Warnings and Errors An aspect may specify that a particular join point should never be reached. * `declare error: Pointcut : String ;` * `declare warning: Pointcut : String ;` If the compiler determines that a join point in `Pointcut` could possibly be reached, then it will signal either an error or warning, as declared, using the `String` for its message. ==== Softened exceptions An aspect may specify that a particular kind of exception, if thrown at a join point, should bypass Java's usual static exception checking system and instead be thrown as a `org.aspectj.lang.SoftException`, which is subtype of `RuntimeException` and thus does not need to be declared. * `declare soft: Type : Pointcut ;` For example, the aspect [source, java] .... aspect A { declare soft: Exception: execution(void main(String[] args)); } .... Would, at the execution join point, catch any `Exception` and rethrow a `org.aspectj.lang.SoftException` containing original exception. This is similar to what the following advice would do [source, java] .... aspect A { void around() execution(void main(String[] args)) { try { proceed(); } catch (Exception e) { throw new org.aspectj.lang.SoftException(e); } } } .... except, in addition to wrapping the exception, it also affects Java's static exception checking mechanism. Like advice, the declare soft form has no effect in an abstract aspect that is not extended by a concreate aspect. So the following code will not compile unless it is compiled with an extending concrete aspect: [source, java] .... abstract aspect A { abstract pointcut softeningPC(); before() : softeningPC() { Class.forName("FooClass"); // error: uncaught ClassNotFoundException } declare soft : ClassNotFoundException : call(* Class.*(..)); } .... [[advice-precedence-cross]] ==== Advice Precedence An aspect may declare a precedence relationship between concrete aspects with the `declare precedence` form: * `declare precedence : TypePatternList ;` This signifies that if any join point has advice from two concrete aspects matched by some pattern in `TypePatternList`, then the precedence of the advice will be the order of in the list. In `TypePatternList`, the wildcard `*` can appear at most once, and it means "any type not matched by any other pattern in the list". For example, the constraints that (1) aspects that have Security as part of their name should have precedence over all other aspects, and (2) the Logging aspect (and any aspect that extends it) should have precedence over all non-security aspects, can be expressed by: [source, java] .... declare precedence: *..*Security*, Logging+, *; .... For another example, the `CountEntry` aspect might want to count the entry to methods in the current package accepting a Type object as its first argument. However, it should count all entries, even those that the aspect `DisallowNulls` causes to throw exceptions. This can be accomplished by stating that `CountEntry` has precedence over `DisallowNulls`. This declaration could be in either aspect, or in another, ordering aspect: [source, java] .... aspect Ordering { declare precedence: CountEntry, DisallowNulls; } aspect DisallowNulls { pointcut allTypeMethods(Type obj): call(* *(..)) && args(obj, ..); before(Type obj): allTypeMethods(obj) { if (obj == null) throw new RuntimeException(); } } aspect CountEntry { pointcut allTypeMethods(Type obj): call(* *(..)) && args(obj, ..); static int count = 0; before(): allTypeMethods(Type) { count++; } } .... ===== Various cycles It is an error for any aspect to be matched by more than one TypePattern in a single decare precedence, so: [source, java] .... declare precedence: A, B, A ; // error .... However, multiple declare precedence forms may legally have this kind of circularity. For example, each of these declare precedence is perfectly legal: [source, java] .... declare precedence: B, A; declare precedence: A, B; .... And a system in which both constraints are active may also be legal, so long as advice from `A` and `B` don't share a join point. So this is an idiom that can be used to enforce that `A` and `B` are strongly independent. ===== Applies to concrete aspects Consider the following library aspects: [source, java] .... abstract aspect Logging { abstract pointcut logged(); before(): logged() { System.err.println("thisJoinPoint: " + thisJoinPoint); } } abstract aspect MyProfiling { abstract pointcut profiled(); Object around(): profiled() { long beforeTime = System.currentTimeMillis(); try { return proceed(); } finally { long afterTime = System.currentTimeMillis(); addToProfile(thisJoinPointStaticPart, afterTime - beforeTime); } } abstract void addToProfile( org.aspectj.JoinPoint.StaticPart jp, long elapsed ); } .... In order to use either aspect, they must be extended with concrete aspects, say, MyLogging and MyProfiling. Because advice only applies from concrete aspects, the declare precedence form only matters when declaring precedence with concrete aspects. So [source, java] .... declare precedence: Logging, Profiling; .... has no effect, but both [source, java] .... declare precedence: MyLogging, MyProfiling; declare precedence: Logging+, Profiling+; .... are meaningful. ==== Statically determinable pointcuts Pointcuts that appear inside of `declare` forms have certain restrictions. Like other pointcuts, these pick out join points, but they do so in a way that is statically determinable. Consequently, such pointcuts may not include, directly or indirectly (through user-defined pointcut declarations) pointcuts that discriminate based on dynamic (runtime) context. Therefore, such pointcuts may not be defined in terms of * `cflow` * `cflowbelow` * `this` * `target` * `args` * `if` all of which can discriminate on runtime information. [[semantics-aspects]] === Aspects An aspect is a crosscutting type defined by the `aspect` declaration. ==== Aspect Declaration The `aspect` declaration is similar to the `class` declaration in that it defines a type and an implementation for that type. It differs in a number of ways: ===== Aspect implementation can cut across other types In addition to normal Java class declarations such as methods and fields, aspect declarations can include AspectJ declarations such as advice, pointcuts, and inter-type declarations. Thus, aspects contain implementation declarations that can can cut across other types (including those defined by other aspect declarations). ===== Aspects are not directly instantiated Aspects are not directly instantiated with a new expression, with cloning, or with serialization. Aspects may have one constructor definition, but if so it must be of a constructor taking no arguments and throwing no checked exceptions. ===== Nested aspects must be `static` Aspects may be defined either at the package level, or as a `static` nested aspect -- that is, a `static` member of a class, interface, or aspect. If it is not at the package level, the aspect _must_ be defined with the `static` keyword. Local and anonymous aspects are not allowed. ==== Aspect Extension To support abstraction and composition of crosscutting concerns, aspects can be extended in much the same way that classes can. Aspect extension adds some new rules, though. ===== Aspects may extend classes and implement interfaces An aspect, abstract or concrete, may extend a class and may implement a set of interfaces. Extending a class does not provide the ability to instantiate the aspect with a new expression: The aspect may still only define a null constructor. ===== Classes may not extend aspects It is an error for a class to extend or implement an aspect. ===== Aspects extending aspects Aspects may extend other aspects, in which case not only are fields and methods inherited but so are pointcuts. However, aspects may only extend abstract aspects. It is an error for a concrete aspect to extend another concrete aspect. ==== Aspect instantiation Unlike class expressions, aspects are not instantiated with `new` expressions. Rather, aspect instances are automatically created to cut across programs. A program can get a reference to an aspect instance using the static method `aspectOf(..)`. Because advice only runs in the context of an aspect instance, aspect instantiation indirectly controls when advice runs. The criteria used to determine how an aspect is instantiated is inherited from its parent aspect. If the aspect has no parent aspect, then by default the aspect is a singleton aspect. How an aspect is instantiated controls the form of the `aspectOf(..)` method defined on the concrete aspect class. ===== Singleton Aspects * `aspect Id { ... }` * `aspect Id issingleton() { ... }` By default (or by using the modifier `issingleton()`) an aspect has exactly one instance that cuts across the entire program. That instance is available at any time during program execution from the static method `aspectOf()` automatically defined on all concrete aspects -- so, in the above examples, `A.aspectOf()` will return ``A``'s instance. This aspect instance is created as the aspect's classfile is loaded. Because the an instance of the aspect exists at all join points in the running of a program (once its class is loaded), its advice will have a chance to run at all such join points. (In actuality, one instance of the aspect `A` is made for each version of the aspect `A`, so there will be one instantiation for each time `A` is loaded by a different classloader.) ===== Per-object aspects * `aspect Id perthis( Pointcut ) { ... }` * `aspect Id pertarget( Pointcut ) { ... }` If an aspect `A` is defined `perthis(Pointcut)`, then one object of type `A` is created for every object that is the executing object (i.e., `this`) at any of the join points picked out by `Pointcut`. The advice defined in `A` will run only at a join point where the currently executing object has been associated with an instance of `A`. Similarly, if an aspect `A` is defined `pertarget(Pointcut)`, then one object of type `A` is created for every object that is the target object of the join points picked out by `Pointcut`. The advice defined in `A` will run only at a join point where the target object has been associated with an instance of `A`. In either case, the static method call `A.aspectOf(Object)` can be used to get the aspect instance (of type `A`) registered with the object. Each aspect instance is created as early as possible, but not before reaching a join point picked out by `Pointcut` where there is no associated aspect of type `A`. Both `perthis` and `pertarget` aspects may be affected by code the AspectJ compiler controls, as discussed in the xref:implementation.adoc#implementation[Implementation Notes] appendix. ===== Per-control-flow aspects * `aspect Id percflow( Pointcut ) { ... }` * `aspect Id percflowbelow( Pointcut ) { ... }` If an aspect `A` is defined `percflow(Pointcut)` or `percflowbelow(Pointcut)`, then one object of type `A` is created for each flow of control of the join points picked out by `Pointcut`, either as the flow of control is entered, or below the flow of control, respectively. The advice defined in `A` may run at any join point in or under that control flow. During each such flow of control, the static method `A.aspectOf()` will return an object of type `A`. An instance of the aspect is created upon entry into each such control flow. ===== Aspect instantiation and advice All advice runs in the context of an aspect instance, but it is possible to write a piece of advice with a pointcut that picks out a join point that must occur before asopect instantiation. For example: [source, java] .... public class Client { public static void main(String[] args) { Client c = new Client(); } } aspect Watchcall { pointcut myConstructor(): execution(new(..)); before(): myConstructor() { System.err.println("Entering Constructor"); } } .... The before advice should run before the execution of all constructors in the system. It must run in the context of an instance of the Watchcall aspect. The only way to get such an instance is to have Watchcall's default constructor execute. But before that executes, we need to run the before advice... There is no general way to detect these kinds of circularities at compile time. If advice runs before its aspect is instantiated, AspectJ will throw a xref:../api/org/aspectj/lang/NoAspectBoundException.html[`org.aspectj.lang.NoAspectBoundException`]. ==== Aspect privilege * `privileged aspect Id { ... }` Code written in aspects is subject to the same access control rules as Java code when referring to members of classes or aspects. So, for example, code written in an aspect may not refer to members with default (package-protected) visibility unless the aspect is defined in the same package. While these restrictions are suitable for many aspects, there may be some aspects in which advice or inter-type members needs to access private or protected resources of other types. To allow this, aspects may be declared `privileged`. Code in priviliged aspects has access to all members, even private ones. [source, java] .... class C { private int i = 0; void incI(int x) { i = i+x; } } privileged aspect A { static final int MAX = 1000; before(int x, C c): call(void C.incI(int)) && target(c) && args(x) { if (c.i+x > MAX) throw new RuntimeException(); } } .... In this case, if `A` had not been declared `privileged`, the field reference `c.i` would have resulted in an error signaled by the compiler. If a privileged aspect can access multiple versions of a particular member, then those that it could see if it were not privileged take precedence. For example, in the code [source, java] .... class C { private int i = 0; void foo() { } } privileged aspect A { private int C.i = 999; before(C c): call(void C.foo()) target(c) { System.out.println(c.i); } } .... ``A``'s private inter-type field `C.i`, initially bound to 999, will be referenced in the body of the advice in preference to ``C``'s privately declared field, since `A` would have access to its own inter-type fields even if it were not privileged. Note that a privileged aspect can access private inter-type declarations made by other aspects, since they are simply considered private members of that other aspect.