1. Aspect-Oriented Programming with AspectJ™
AspectJ Tutorial
Aspect-Oriented Programming with AspectJ™
the AspectJ.org team
Xerox PARC
Bill Griswold, Erik Hilsdale, Jim Hugunin, Mik Kersten, Gregor Kiczales, Jeffrey Palm
partially funded by DARPA under contract F C0246
AspectJ Tutorial

2. this tutorial is about... using AOP and AspectJ to:
AspectJ Tutorial
this tutorial is about...
using AOP and AspectJ to:
improve the modularity of crosscutting concerns
design modularity
source code modularity
development process
aspects are two things:
concerns that crosscut [design level]
a programming construct [implementation level]
enables crosscutting concerns to be captured in modular units
AspectJ is:
is an aspect-oriented extension to Java™ that supports general-purpose aspect-oriented programming
AspectJ Tutorial

3. problems like… logging is not modularized
where is logging in org.apache.tomcat
red shows lines of code that handle logging
not in just one place
not even in a small number of places
AspectJ Tutorial

4. StandardSessionManager
problems like…
session expiration is not modularized /* *
* ====================================================================
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* reserved.
* Copyright (c) 1999 The Apache Software Foundation. All rights
* The Apache Software License, Version 1.1
* 1. Redistributions of source code must retain the above copyright
* are met:
* the documentation and/or other materials provided with the
* distribution.
* notice, this list of conditions and the following disclaimer in
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer.
itself,
* if and wherever such third-party acknowlegements normally appear.
* Apache Software Foundation (
* Alternately, this acknowlegement may appear in the software
* "This product includes software developed by the
* 3. The end-user documentation included with the redistribution, if
* any, must include the following acknowlegement:
* Foundation" must not be used to endorse or promote products
* 4. The names "The Jakarta Project", "Tomcat", and "Apache Software
* nor may "Apache" appear in their names without prior written
* permission of the Apache Group.
* 5. Products derived from this software may not be called "Apache"
* permission, please contact
derived
* from this software without prior written permission. For written
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
* ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
* THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESSED OR IMPLIED
* DISCLAIMED. IN NO EVENT SHALL THE APACHE SOFTWARE FOUNDATION OR
* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* This software consists of voluntary contributions made by many
* SUCH DAMAGE.
* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
* USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
* ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
* information on the Apache Software Foundation, please see
* individuals on behalf of the Apache Software Foundation. For more
* [Additional notices, if required by prior licensing conditions] */
* <
package org.apache.tomcat.session;
import java.net.*;
import java.util.*;
import java.io.*;
import org.apache.tomcat.util.StringManager;
import org.apache.tomcat.core.*;
* Core implementation of an application level session
James Duncan Davidson
/**
import javax.servlet.http.*;
import javax.servlet.*;
James Todd
Jason Hunter
private String id;
private Hashtable values = new Hashtable();
private StringManager sm =
public class ApplicationSession implements HttpSession {
StringManager.getManager("org.apache.tomcat.session");
private ServerSession serverSession;
private boolean valid = true;
private int inactiveInterval = -1;
private long lastAccessed = creationTime;
private Context context;
private long creationTime = System.currentTimeMillis();;
private long thisAccessTime = creationTime;
this.inactiveInterval = context.getSessionTimeOut();
this.context = context;
this.id = id;
Context context) {
ApplicationSession(String id, ServerSession serverSession,
this.serverSession = serverSession; }
this.inactiveInterval *= 60;
if (this.inactiveInterval != -1) {
return serverSession;
ServerSession getServerSession() {
* inactivities can be dealt with accordingly.
* Called by context when request comes in so that accesses and
void accessed() {
thisAccessTime = System.currentTimeMillis();
lastAccessed = thisAccessTime;
// set last accessed to thisAccessTime as it will be left over
// from the previous access
int thisInterval =
if (inactiveInterval != -1) {
void validate() {
validate();
invalidate();
if (thisInterval > inactiveInterval) {
(int)(System.currentTimeMillis() - lastAccessed) / 1000;
// if we have an inactive interval, check to see if we've exceeded it
if (valid) {
// HTTP SESSION IMPLEMENTATION METHODS
String msg = sm.getString("applicationSession.session.ise");
public String getId() {
throw new IllegalStateException(msg);
} else {
return id;
return creationTime;
public long getCreationTime() {
public long getLastAccessedTime() {
return new SessionContextImpl();
public HttpSessionContext getSessionContext() {
return lastAccessed;
public void invalidate() {
// remove everything in the session
serverSession.removeApplicationSession(context);
removeValue(name);
String name = (String)enum.nextElement();
Enumeration enum = values.keys();
while (enum.hasMoreElements()) {
public boolean isNew() {
valid = false;
if (! valid) {
return true;
return false;
if (thisAccessTime == creationTime) {
public void setAttribute(String name, Object value) {
setAttribute(name, value);
public void putValue(String name, Object value) {
removeValue(name); // remove any existing binding
throw new IllegalArgumentException(msg);
String msg = sm.getString("applicationSession.value.iae");
if (name == null) {
new HttpSessionBindingEvent(this, name);
((HttpSessionBindingListener)value).valueBound(e);
HttpSessionBindingEvent e =
if (value != null && value instanceof HttpSessionBindingListener) {
values.put(name, value);
public Object getAttribute(String name) {
return getAttribute(name);
public Object getValue(String name) {
return values.get(name);
public String[] getValueNames() {
names.addElement(e.nextElement());
while (e.hasMoreElements()) {
Vector names = new Vector();
Enumeration e = getAttributeNames();
names.copyInto(valueNames);
String[] valueNames = new String[names.size()];
public Enumeration getAttributeNames() {
return valueNames;
return (Enumeration)valuesClone.keys();
Hashtable valuesClone = (Hashtable)values.clone();
public void removeAttribute(String name) {
removeAttribute(name);
public void removeValue(String name) {
Object o = values.get(name);
((HttpSessionBindingListener)o).valueUnbound(e);
new HttpSessionBindingEvent(this,name);
if (o instanceof HttpSessionBindingListener) {
public void setMaxInactiveInterval(int interval) {
values.remove(name);
public int getMaxInactiveInterval() {
inactiveInterval = interval;
return inactiveInterval; //
ApplicationSession
import java.io.IOException;
import java.util.Vector;
import javax.servlet.ServletException;
import java.util.Hashtable;
import java.util.Enumeration;
import java.io.ObjectInputStream;
import java.io.ObjectOutputStream;
import java.io.Serializable;
import javax.servlet.http.HttpSession;
import javax.servlet.http.HttpSessionContext;
import org.apache.tomcat.catalina.*;
import javax.servlet.http.HttpSessionBindingListener;
import javax.servlet.http.HttpSessionBindingEvent;
* serializable, so that it can be stored in persistent storage or transferred
* Standard implementation of the <b>Session</b> interface. This object is
* internal (Session) and application level (HttpSession) view of the session.
* <b>IMPLEMENTATION NOTE</b>: An instance of this class represents both the
* <p>
* to a different JVM for distributable session support.
* HttpSession view of this instance back to a Session view.
* of the <code>org.apache.tomcat.session</code> package cannot cast an
* However, because the class itself is not declared public, Java logic outside
$Revision: 1.2 $ $Date: 2000/05/15 17:54:10 $
Craig R. McClanahan
implements HttpSession, Session {
final class StandardSession
// Constructors
manager The manager with which this Session is associated
* Construct a new Session associated with the specified Manager.
this.manager = manager;
super();
public StandardSession(Manager manager) {
// Instance Variables
private Hashtable attributes = new Hashtable();
* The collection of user data attributes associated with this Session.
* The time this session was created, in milliseconds since midnight,
private long creationTime = 0L;
* January 1, 1970 GMT.
private String id = null;
* The session identifier of this Session.
private static final String info = "StandardSession/1.0";
* Descriptive information describing this Session implementation.
private long lastAccessedTime = creationTime;
* The last accessed time for this Session.
private Manager manager = null;
* The Manager with which this Session is associated.
* the servlet container may invalidate this session. A negative time
* The maximum time interval, in seconds, between client requests before
* indicates that the session should never time out.
private boolean isNew = true;
* Flag indicating whether this session is new or not.
private int maxInactiveInterval = -1;
private boolean isValid = false;
* Flag indicating whether this session is valid or not.
* The string manager for this package.
private static HttpSessionContext sessionContext = null;
* The HTTP session context associated with this session.
private long thisAccessedTime = creationTime;
* The current accessed time for this session.
// Session Properties
* Set the creation time for this session. This method is called by the
public void setCreationTime(long time) {
time The new creation time
* Manager when an existing Session instance is reused.
this.thisAccessedTime = time;
this.lastAccessedTime = time;
this.creationTime = time;
* Return the session identifier for this session.
return (this.id);
(manager instanceof ManagerBase))
if ((this.id != null) && (manager != null) &&
public void setId(String id) {
* Set the session identifier for this session.
id The new session identifier
((ManagerBase) manager).remove(this);
((ManagerBase) manager).add(this);
if ((manager != null) && (manager instanceof ManagerBase))
* <code><description>/<version></code>.
* Return descriptive information about this Session implementation and
* the corresponding version number, in the format
return (this.info);
public String getInfo() {
* a value associated with the session, do not affect the access time.
* GMT. Actions that your application takes, such as getting or setting
* session, as the number of milliseconds since midnight, January 1, 1970
* Return the last time the client sent a request associated with this
return (this.lastAccessedTime);
return (this.manager);
public Manager getManager() {
* Return the Manager within which this Session is valid.
public void setManager(Manager manager) {
manager The new Manager
* Set the Manager within which this Session is valid.
* Return the maximum time interval, in seconds, between client requests
IllegalStateException if this method is called on
* time indicates that the session should never time out.
* before the servlet container will invalidate the session. A negative
return (this.maxInactiveInterval);
* an invalidated session
* Set the maximum time interval, in seconds, between client requests
interval The new maximum interval
this.maxInactiveInterval = interval;
public HttpSession getSession() {
* is the facade.
* Return the <code>HttpSession</code> for which this object
return ((HttpSession) this);
// Session Public Methods
* session, even if the application does not reference it.
* should be called by the context when a request comes in for a particular
* Update the accessed time information for this session. This method
this.isNew=false;
this.thisAccessedTime = System.currentTimeMillis();
this.lastAccessedTime = this.thisAccessedTime;
public void access() {
* without triggering an exception if the session has already expired.
* Perform the internal processing required to invalidate this session,
// Remove this session from our manager's active sessions
public void expire() {
while (attrs.hasMoreElements()) {
Enumeration attrs = getAttributeNames();
Vector results = new Vector();
// Unbind any objects associated with this session
while (names.hasMoreElements()) {
String name = (String) names.nextElement();
Enumeration names = results.elements();
results.addElement(attr);
String attr = (String) attrs.nextElement();
setValid(false);
// Mark this session as invalid
public void recycle() {
* preparation for reuse of this object.
* Release all object references, and initialize instance variables, in
// Reset the instance variables associated with this Session
manager = null;
isNew = true;
maxInactiveInterval = -1;
lastAccessedTime = 0L;
attributes.clear();
creationTime = 0L;
id = null;
isValid = false;
((ManagerBase) manager).recycle(this);
// Tell our Manager that this Session has been recycled
// Session Package Methods
boolean isValid() {
* Return the <code>isValid</code> flag for this session.
return (this.isValid);
void setNew(boolean isNew) {
isNew The new value for the <code>isNew</code> flag
* Set the <code>isNew</code> flag for this session.
this.isNew = isNew;
isValid The new value for the <code>isValid</code> flag
* Set the <code>isValid</code> flag for this session.
this.isValid = isValid;
void setValid(boolean isValid) {
// HttpSession Properties
IllegalStateException if this method is called on an
* Return the time when this session was created, in milliseconds since
* midnight, January 1, 1970 GMT.
* invalidated session
return (this.creationTime);
* Return the session context with which this session is associated.
* Java Servlet API.
* replacement. It will be removed in a future version of the
As of Version 2.1, this method is deprecated and has no
return (sessionContext);
sessionContext = new StandardSessionContext();
if (sessionContext == null)
// HttpSession Public Methods
* <code>null</code> if no object is bound with that name.
* Return the object bound with the specified name in this session, or
name Name of the attribute to be returned
return (attributes.get(name));
* Return an <code>Enumeration</code> of <code>String</code> objects
* containing the names of the objects bound to this session.
return (attributes.keys());
name Name of the value to be returned
* <code>getAttribute()</code>
As of Version 2.2, this method is replaced by
return (getAttribute(name));
* are no such objects, a zero-length array is returned.
* Return the set of names of objects bound to this session. If there
* <code>getAttributeNames()</code>
names[i] = (String) results.elementAt(i);
return (names);
for (int i = 0; i < names.length; i++)
String names[] = new String[results.size()];
* Invalidates this session and unbinds any objects bound to it.
expire();
// Cause this session to expire
* session, or if the client chooses not to join the session. For
* Return <code>true</code> if the client does not yet know about the
* request.
* has disabled the use of cookies, then a session would be new on each
* example, if the server used only cookie-based sessions, and the client
return (this.isNew);
* Bind an object to this session, using the specified name. If an object
* replaced.
* of the same name is already bound to this session, the object is
name Name to which the object is bound, cannot be null
value Object to be bound, cannot be null
* <code>valueBound()</code> on the object.
* After this method executes, and if the object implements
* <code>HttpSessionBindingListener</code>, the container calls
* <code>setAttribute()</code>
* does nothing.
* the session does not have an object bound with this name, this method
* Remove the object bound with the specified name from this session. If
name Name of the object to remove from this session.
* <code>valueUnbound()</code> on the object.
Object object = attributes.get(name);
synchronized (attributes) {
((HttpSessionBindingListener) object).valueUnbound
(new HttpSessionBindingEvent((HttpSession) this, name));
if (object instanceof HttpSessionBindingListener) {
// System.out.println( "Removing attribute " + name );
return;
attributes.remove(name);
if (object == null)
* <code>removeAttribute()</code>
IllegalArgumentException if an attempt is made to add a
* non-serializable object in an environment marked distributable.
throw new IllegalArgumentException
(sm.getString("standardSession.setAttribute.iae"));
!(value instanceof Serializable))
if ((manager != null) && manager.getDistributable() &&
((HttpSessionBindingListener) value).valueBound
if (value instanceof HttpSessionBindingListener)
attributes.put(name, value);
// HttpSession Private Methods
* is not restored by this method, and must be set explicitly.
* <b>IMPLEMENTATION NOTE</b>: The reference to the owning Manager
* object input stream.
* Read a serialized version of this session object from the specified
ClassNotFoundException if an unknown class is specified
IOException if an input/output error occurs
stream The input stream to read from
id = (String) stream.readObject();
lastAccessedTime = ((Long) stream.readObject()).longValue();
creationTime = ((Long) stream.readObject()).longValue();
// Deserialize the scalar instance variables (except Manager)
private void readObject(ObjectInputStream stream)
throws ClassNotFoundException, IOException {
isNew = ((Boolean) stream.readObject()).booleanValue();
maxInactiveInterval = ((Integer) stream.readObject()).intValue();
String name = (String) stream.readObject();
Object value = (Object) stream.readObject();
for (int i = 0; i < n; i++) {
int n = ((Integer) stream.readObject()).intValue();
// Deserialize the attribute count and attribute values
isValid = ((Boolean) stream.readObject()).booleanValue();
* object output stream.
* Write a serialized version of this session object to the specified
* explicitly.
* <b>IMPLEMENTATION NOTE</b>: Any attribute that is not Serializable
* <code>readObject()</code>, you must set the associated Manager
* in the serialized representation of this Session. After calling
* <b>IMPLEMENTATION NOTE</b>: The owning Manager will not be stored
stream The output stream to write to
* Manager is set to <code>true</code>.
* be sure the <code>distributable</code> property of our associated
* will be silently ignored. If you do not want any such attributes,
stream.writeObject(id);
stream.writeObject(new Long(creationTime));
private void writeObject(ObjectOutputStream stream) throws IOException {
stream.writeObject(new Boolean(isValid));
stream.writeObject(new Integer(maxInactiveInterval));
stream.writeObject(new Long(lastAccessedTime));
// Write the scalar instance variables (except Manager)
// Accumulate the names of serializable attributes
stream.writeObject(new Boolean(isNew));
if (value instanceof Serializable)
Object value = attributes.get(attr);
stream.writeObject(new Integer(results.size()));
// Serialize the attribute count and the attribute values
stream.writeObject(attributes.get(name));
stream.writeObject(name);
Enumeration getAttributeNames() |
Object getAttribute(String) |
crosscut invalidate(StandardSession s): s & (int getMaxInactiveInterval() |
long getCreationTime() |
String[] getValueNames() |
static advice(StandardSession s): invalidate(s) {
if (!s.isValid())
before {
void setAttribute(String, Object));
void removeAttribute(String) |
void invalidate() |
boolean isNew() |
+ ".ise"));
+ thisJoinPoint.methodName
throw new IllegalStateException
(s.sm.getString("standardSession."
// Private Class
* This class is a dummy implementation of the <code>HttpSessionContext</code>
* when <code>HttpSession.getSessionContext()</code> is called.
* interface, to conform to the requirement that such an object be returned
* interface will be removed in a future version of this API.
As of Java Servlet API 2.1 with no replacement. The
* Return the session identifiers of all sessions defined
* within this context.
private Vector dummy = new Vector();
final class StandardSessionContext implements HttpSessionContext {
public Enumeration getIds() {
* and will be removed in a future version of the API.
As of Java Servlet API 2.1 with no replacement.
* This method must return an empty <code>Enumeration</code>
return (dummy.elements());
id Session identifier for which to look up a session
* Return the <code>HttpSession</code> associated with the
* specified session identifier.
return (null);
public HttpSession getSession(String id) {
* future version of the API.
* This method must return null and will be removed in a
StandardSession
import javax.servlet.http.Cookie;
import org.w3c.dom.Node;
import org.w3c.dom.NamedNodeMap;
* an optional, configurable, maximum number of active sessions allowed.
* no session persistence or distributable capabilities, but does support
* Standard implementation of the <b>Manager</b> interface that provides
* checkInterval="60" maxActiveSessions="-1"
* maxInactiveInterval="-1" />
* <Manager className="org.apache.tomcat.session.StandardManager"
* <code>
* in the following format:
* Lifecycle configuration of this component assumes an XML node
* where you can adjust the following parameters, with default values
* </code>
* be active at once, or -1 for no limit. [-1]
* <li><b>maxInactiveInterval</b> - The default maximum number of seconds of
* inactivity before which the servlet container is allowed to time out
* <li><b>maxActiveSessions</b> - The maximum number of sessions allowed to
* thread checks for expired sessions. [60]
* <li><b>checkInterval</b> - The interval (in seconds) between background
* in square brackets:
* <ul>
* </ul>
* descriptor, if any. [-1]
* the default session timeout specified in the web application deployment
* a session, or -1 for no limit. This value should be overridden from
implements Lifecycle, Runnable {
extends ManagerBase
public final class StandardManager
$Revision: $ $Date: 2000/05/02 21:28:30 $
* The interval (in seconds) between checks for expired sessions.
private boolean configured = false;
* Has this component been configured yet?
private int checkInterval = 60;
private static final String info = "StandardManager/1.0";
* The descriptive information about this implementation.
protected int maxActiveSessions = -1;
* The maximum number of active Sessions allowed, or -1 for no limit.
private boolean started = false;
* Has this component been started yet?
private Thread thread = null;
* The background thread.
* The background thread completion semaphore.
private boolean threadDone = false;
* Name to register for the background thread.
// Properties
private String threadName = "StandardManager";
return (this.checkInterval);
public int getCheckInterval() {
* Return the check interval (in seconds) for this Manager.
checkInterval The new check interval
* Set the check interval (in seconds) for this Manager.
public void setCheckInterval(int checkInterval) {
this.checkInterval = checkInterval;
* Return descriptive information about this Manager implementation and
public int getMaxActiveSessions() {
* no limit.
* Return the maximum number of active Sessions allowed, or -1 for
return (this.maxActiveSessions);
* Set the maximum number of actives Sessions allowed, or -1 for
this.maxActiveSessions = max;
public void setMaxActiveSessions(int max) {
max The new maximum number of sessions
// Public Methods
* method of the returned session. If a new session cannot be created
* for any reason, return <code>null</code>.
* id will be assigned by this method, and available via the getId()
* settings specified by this Manager's properties. The session
* Construct and return a new session object, based on the default
if ((maxActiveSessions >= 0) &&
(sessions.size() >= maxActiveSessions))
public Session createSession() {
* instantiated for any reason
IllegalStateException if a new session cannot be
return (super.createSession());
(sm.getString("standardManager.createSession.ise"));
// Lifecycle Methods
* Configure this component, based on the specified configuration
IllegalStateException if this component has already been
* (<B>FIXME: What object type should this really be?)
parameters Configuration parameters for this component
* parameters. This method should be called immediately after the
* component instance is created, and before <code>start()</code>
* is called.
LifecycleException if this component detects a fatal error
* configured and/or started
if (configured)
// Validate and update our current component state
throws LifecycleException {
* in the configuration parameters it was given
public void configure(Node parameters)
configured = true;
(sm.getString("standardManager.alreadyConfigured"));
throw new LifecycleException
NamedNodeMap attributes = parameters.getAttributes();
if (!("Manager".equals(parameters.getNodeName())))
// Parse and process our configuration parameters
if (parameters == null)
Node node = null;
} catch (Throwable t) {
; // XXX - Throw exception?
setCheckInterval(Integer.parseInt(node.getNodeValue()));
try {
if (node != null) {
node = attributes.getNamedItem("checkInterval");
setMaxActiveSessions(Integer.parseInt(node.getNodeValue()));
node = attributes.getNamedItem("maxActiveSessions");
setMaxInactiveInterval(Integer.parseInt(node.getNodeValue()));
node = attributes.getNamedItem("maxInactiveInterval");
IllegalStateException if this component has not yet been
* configured (if required for this component)
* and before any of the public methods of the component are utilized.
* component. This method should be called after <code>configure()</code>,
* Prepare for the beginning of active use of the public methods of this
public void start() throws LifecycleException {
* that prevents this component from being used
* started
(sm.getString("standardManager.notConfigured"));
if (!configured)
// Start the background reaper thread
threadStart();
started = true;
(sm.getString("standardManager.alreadyStarted"));
if (started)
* component. This method should be the last one called on a given
* Gracefully terminate the active use of the public methods of this
* been stopped
* that needs to be reported
IllegalStateException if this component has already
* instance of this component.
IllegalStateException if this component has not been started
public void stop() throws LifecycleException {
// Stop the background reaper thread
threadStop();
started = false;
if (!started)
(sm.getString("standardManager.notStarted"));
// Expire all active sessions
continue;
session.expire();
if (!session.isValid())
StandardSession session = (StandardSession) sessions[i];
Session sessions[] = findSessions();
for (int i = 0; i < sessions.length; i++) {
// Private Methods
* Invalidate all sessions that have expired.
long timeNow = System.currentTimeMillis();
private void processExpires() {
(int) ((timeNow - session.getLastAccessedTime()) / 1000L);
if (timeIdle >= maxInactiveInterval)
int timeIdle = // Truncate, do not round up
if (maxInactiveInterval < 0)
int maxInactiveInterval = session.getMaxInactiveInterval();
private void threadSleep() {
* property.
* Sleep for the duration specified by the <code>checkInterval</code> ;
} catch (InterruptedException e) {
Thread.sleep(checkInterval * 1000L);
if (thread != null)
private void threadStart() {
* Start the background thread that will periodically check for
* session timeouts.
thread.start();
thread.setDaemon(true);
thread = new Thread(this, threadName);
threadDone = false;
private void threadStop() {
* Stop the background thread that is periodically checking for
if (thread == null)
thread.join();
thread.interrupt();
threadDone = true;
thread = null;
// Background Thread
* The background thread that checks for session timeouts and shutdown.
processExpires();
threadSleep();
// Loop until the termination semaphore is set
public void run() {
while (!threadDone) {
StandardManager
StandardSessionManager
import org.apache.tomcat.util.SessionUtil;
import org.apache.tomcat.core.SessionManager;
import org.apache.tomcat.core.Response;
import org.apache.tomcat.core.Context;
import org.apache.tomcat.core.Request;
* XXX - At present, use of <code>StandardManager</code> is hard coded,
* that adapts to the new component-based Manager implementation.
* Specialized implementation of org.apache.tomcat.core.SessionManager
* and lifecycle configuration is not supported.
* the core level. The Tomcat.Next "Manager" interface acts more like a
* collection class, and has minimal knowledge of the detailed request
* paradigm, I would suggest moving the logic implemented here back into
* <b>IMPLEMENTATION NOTE</b>: Once we commit to the new Manager/Session
* a Context to tell the Manager that we create what the default session
* XXX - At present, there is no way (via the SessionManager interface) for
* processing semantics of handling sessions.
* should be.
* timeout for this web application (specified in the deployment descriptor)
implements SessionManager {
public final class StandardSessionManager
* Create a new SessionManager that adapts to the corresponding Manager
manager = new StandardManager();
public StandardSessionManager() {
* implementation.
if (manager instanceof Lifecycle) {
} catch (LifecycleException e) {
throw new IllegalStateException("" + e);
((Lifecycle) manager).start();
((Lifecycle) manager).configure(null);
* The Manager implementation we are actually using.
* Mark the specified session's last accessed time. This should be
public void accessed(Context ctx, Request req, String id) {
session The session to be marked
* called for each request by a RequestInterceptor.
HttpSession session=findSession(ctx, id);
// cache the HttpSession - avoid another find
((Session) session).access();
if (session instanceof Session)
if( session == null) return;
req.setSession( session );
Session session = manager.findSession(id);
public HttpSession findSession( Context ctx, String id ) {
// XXX should we throw exception or just return null ??
} catch (IOException e) {
if(session!=null)
return session.getSession();
return manager.createSession().getSession();
public HttpSession createSession(Context ctx) {
* Remove all sessions because our associated Context is being shut down.
ctx The context that is being shut down
// contexts, we just want to remove the sessions of ctx!
// The manager will still run after that ( i.e. keep database
// XXX XXX a manager may be shared by multiple
public void removeSessions(Context ctx) {
// connection open
((Lifecycle) manager).stop();
* Used by context to configure the session manager's inactivity timeout.
* descriptor (web.xml). This method lets the Context conforgure the
* Context on the other hand has it's timeout set by the deployment
* The SessionManager may have some default session time out, the
public void setSessionTimeOut(int minutes) {
minutes The session inactivity timeout in minutes.
* session manager according to this value.
manager.setMaxInactiveInterval(minutes * 60);
if(-1 != minutes) {
// The manager works with seconds...
ServerSessionManager
import org.apache.tomcat.util.*;
public class ServerSessionManager implements SessionManager {
manager = new ServerSessionManager();
static {
protected int inactiveInterval = -1;
private static ServerSessionManager manager; // = new ServerSessionManager();
private Reaper reaper;
private Hashtable sessions = new Hashtable();
return manager;
public static ServerSessionManager getManager() {
reaper.start();
reaper.setServerSessionManager(this);
private ServerSessionManager() {
reaper = Reaper.getReaper();
if( apS==null) return;
ServerSession servS=apS.getServerSession();
ApplicationSession apS=(ApplicationSession)findSession( ctx, id);
public void accessed( Context ctx, Request req, String id ) {
req.setSession( apS );
// cache it - no need to compute it again
apS.accessed();
servS.accessed();
if(-1 != inactiveInterval) {
session.setMaxInactiveInterval(inactiveInterval);
sessions.put(sessionId, session);
ServerSession session = new ServerSession(sessionId);
String sessionId = SessionIdGenerator.generateId();
if(sSession==null) return null;
return sSession.getApplicationSession(ctx, false);
public HttpSession findSession(Context ctx, String id) {
return session.getApplicationSession( ctx, true );
ServerSession sSession=(ServerSession)sessions.get(id);
// XXX
Enumeration enum = sessions.keys();
synchronized void reap() {
// solution for this, but we'll determine something else later.
// sync'd for safty -- no other thread should be getting something
// from this while we are reaping. This isn't the most optimal
session.reap();
Object key = enum.nextElement();
session.validate();
ServerSession session = (ServerSession)sessions.get(key);
String id = session.getId();
synchronized void removeSession(ServerSession session) {
sessions.remove(id);
public void removeSessions(Context context) {
session.invalidate();
appSession.invalidate();
if (appSession != null) {
session.getApplicationSession(context, false);
ApplicationSession appSession =
inactiveInterval = (minutes * 60);
SessionInterceptor
package org.apache.tomcat.request;
* It also marks the session as accessed.
* in the Request.
* Will process the request and determine the session Id, and set it
// GS, separates the session id from the jvm route
static final char SESSIONID_ROUTE_SEP = '.';
public class SessionInterceptor extends BaseInterceptor implements RequestInterceptor {
* add new interceptors for other methods.
* This implementation only handles Cookies sessions, please extend or
public SessionInterceptor() {
int debug=0;
ContextManager cm;
public void setContextManager( ContextManager cm ) {
debug=i;
System.out.println("Set debug to " + i);
public void setDebug( int i ) {
String sessionId = null;
public int requestMap(Request request ) {
this.cm=cm;
if (cookie.getName().equals("JSESSIONID")) {
sessionId = cookie.getValue();
Cookie cookie = cookies[i];
for( int i=0; i<cookies.length; i++ ) {
Cookie cookies[]=request.getCookies(); // assert !=null
request.setRequestedSessionIdFromCookie(true);
if (sessionId!=null){
sessionId=validateSessionId(request, sessionId);
sessionId=request.getRequestURI().substring(foundAt+sig.length());
request.setRequestURI(request.getRequestURI().substring(0, foundAt));
// rewrite URL, do I need to do anything more?
if ((foundAt=request.getRequestURI().indexOf(sig))!=-1){
if( debug>0 ) cm.log(" XXX RURI=" + request.getRequestURI());
String sig=";jsessionid=";
int foundAt=-1;
return 0;
request.setRequestedSessionIdFromURL(true);
* It will also clean up the session from load-balancing strings.
sessionId, or null if not valid
/** Validate and fix the session id. If the session is not valid return null.
// We may still set it and just return session invalid.
// XXX what is the correct behavior if the session is invalid ?
// Separate them ...
// GS, We piggyback the JVM id on top of the session cookie
private String validateSessionId(Request request, String sessionId){
if(idex > 0) {
sessionId = sessionId.substring(0, idex);
int idex = sessionId.lastIndexOf(SESSIONID_ROUTE_SEP);
if( debug>0 ) cm.log(" Orig sessionId " + sessionId );
if (null != sessionId) {
if (sessionId != null && sessionId.length()!=0) {
SessionManager sM = ctx.getSessionManager();
if(null != sM.findSession(ctx, sessionId)) {
sM.accessed(ctx, request, sessionId );
Context ctx=request.getContext();
// cookie. We must check for validity in the current context.
// multiple Session cookies (one for the root
// GS, We are in a problem here, we may actually get
// context and one for the real context... or old session
return sessionId;
if( debug>0 ) cm.log(" Final session id " + sessionId );
request.setRequestedSessionId(sessionId);
return null;
public int beforeBody( Request rrequest, Response response ) {
if( debug>0 ) cm.log("Before Body " + reqSessionId );
if( reqSessionId==null)
String reqSessionId = response.getSessionId();
if(sessionPath.length() == 0) {
sessionPath = "/";
String sessionPath = rrequest.getContext().getPath();
// multiple session cookies can be used, one for each
// context.
// GS, set the path attribute to the cookie. This way
if(null != jvmRoute) {
// GS, piggyback the jvm route on the session id.
reqSessionId = reqSessionId + SESSIONID_ROUTE_SEP + jvmRoute;
String jvmRoute = rrequest.getJvmRoute();
if(!sessionPath.equals("/")) {
cookie.setMaxAge(-1);
reqSessionId);
Cookie cookie = new Cookie("JSESSIONID",
CookieTools.getCookieHeaderValue(cookie));
cookie.setVersion(0);
response.addHeader( CookieTools.getCookieHeaderName(cookie),
cookie.setVersion(1);
cookie.setPath(sessionPath); {
if( ctx.getDebug() > 0 ) ctx.log("Removing sessions from " + ctx );
public void contextShutdown( Context ctx )
throws TomcatException
/** Notification of context shutdown
ctx.getSessionManager().removeSessions(ctx);
ServerSession
* Core implementation of a server session
public class ServerSession {
private Hashtable appSessions = new Hashtable();
ServerSession(String id) {
(ApplicationSession)appSessions.get(context);
public ApplicationSession getApplicationSession(Context context,
boolean create) {
appSession = new ApplicationSession(id, this, context);
appSessions.put(context, appSession);
// sync to ensure valid?
if (appSession == null && create) {
return appSession;
// a new appSession
// inactive interval -- if so, invalidate and create
// make sure that we haven't gone over the end of our
appSessions.remove(context);
void removeApplicationSession(Context context) {
void validate()
// if we have an inactive interval, check to see if
// we've exceeded it
ssm.removeSession(this);
ServerSessionManager ssm =
ServerSessionManager.getManager();
Enumeration enum = appSessions.keys();
synchronized void invalidate() {
(ApplicationSession)appSessions.get(key);
String msg = sm.getString("serverSession.value.iae");
return values.keys();
public Enumeration getValueNames() {
appSession.validate();
AspectJ Tutorial

5. problems like… session tracking is not modularized HTTPRequest
SessionInterceptor
requestMap(request)
beforeBody(req, resp)
...
getCookies()
getRequestURI()(doc)
getSession() getRequestedSessionId()
...
Session
getAttribute(name)
setAttribute(name, val)
invalidate()
...
HTTPResponse
getRequest()
setContentType(contentType)
getOutptutStream() setSessionId(id)
...
Servlet
AspectJ Tutorial

6. the cost of tangled code
redundant code
same fragment of code in many places
difficult to reason about
non-explicit structure
the big picture of the tangling isn’t clear
difficult to change
have to find all the code involved
and be sure to change it consistently
and be sure not to break it by accident
AspectJ Tutorial

7. crosscutting concerns
HTTPRequest
SessionInterceptor
getCookies()
getRequestURI()(doc)
getSession() getRequestedSessionId()
...
requestMap(request)
beforeBody(req, resp)
...
Session
HTTPResponse
getAttribute(name)
setAttribute(name, val)
invalidate()
...
getRequest()
setContentType(contentType)
getOutptutStream() setSessionId(id)
...
Servlet
AspectJ Tutorial

8. the AOP idea crosscutting is inherent in complex systems
AspectJ Tutorial
the AOP idea
aspect-oriented programming
crosscutting is inherent in complex systems
crosscutting concerns
have a clear purpose
have a natural structure
defined set of methods, module boundary crossings, points of resource utilization, lines of dataflow…
so, let’s capture the structure of crosscutting concerns explicitly...
in a modular way
with linguistic and tool support
aspects are
well-modularized crosscutting concerns
here’s where we pick up the crystallization of style and analogy to OOP points
people know this, they deal with it a variety of ways, we can make that explicit
(1) It rests on an assumption that cross-cutting is inherent. That in any sufficiently complex system… I haven’t proven this assumption today, I’ve just reminded you about the problem. But I believe it, and we can talk an EXAMPLES HERE
(2) But that cross-cutting concerns are not random…
AspectJ Tutorial

9. StandardSessionManager StandardSessionManager
AspectJ Tutorial
language support to… /* *
* ====================================================================
* modification, are permitted provided that the following conditions
* Redistribution and use in source and binary forms, with or without
* reserved.
* Copyright (c) 1999 The Apache Software Foundation. All rights
* The Apache Software License, Version 1.1
* are met:
* the documentation and/or other materials provided with the
* distribution.
* notice, this list of conditions and the following disclaimer in
* 2. Redistributions in binary form must reproduce the above copyright
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* Alternately, this acknowlegement may appear in the software
* Apache Software Foundation (
* "This product includes software developed by the
* any, must include the following acknowlegement:
* 3. The end-user documentation included with the redistribution, if
derived
* from this software without prior written permission. For written
* Foundation" must not be used to endorse or promote products
* 4. The names "The Jakarta Project", "Tomcat", and "Apache Software
itself,
* if and wherever such third-party acknowlegements normally appear.
* THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESSED OR IMPLIED
* permission of the Apache Group.
* nor may "Apache" appear in their names without prior written
* 5. Products derived from this software may not be called "Apache"
* permission, please contact
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
* USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
* DISCLAIMED. IN NO EVENT SHALL THE APACHE SOFTWARE FOUNDATION OR
* SUCH DAMAGE.
* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
* information on the Apache Software Foundation, please see
* This software consists of voluntary contributions made by many
* individuals on behalf of the Apache Software Foundation. For more
* [Additional notices, if required by prior licensing conditions] */
* <
package org.apache.tomcat.session;
import javax.servlet.*;
import java.util.*;
import javax.servlet.http.*;
import java.net.*;
import org.apache.tomcat.core.*;
import org.apache.tomcat.util.StringManager;
import java.io.*;
/**
James Todd
Jason Hunter
James Duncan Davidson
* Core implementation of an application level session
public class ApplicationSession implements HttpSession {
private Context context;
private long creationTime = System.currentTimeMillis();;
private ServerSession serverSession;
private String id;
private StringManager sm =
StringManager.getManager("org.apache.tomcat.session");
private Hashtable values = new Hashtable();
this.serverSession = serverSession;
this.context = context;
Context context) {
ApplicationSession(String id, ServerSession serverSession,
private boolean valid = true;
private long thisAccessTime = creationTime;
this.id = id; }
this.inactiveInterval *= 60;
if (this.inactiveInterval != -1) {
this.inactiveInterval = context.getSessionTimeOut();
return serverSession;
ServerSession getServerSession() {
* inactivities can be dealt with accordingly.
* Called by context when request comes in so that accesses and
if (valid) {
// HTTP SESSION IMPLEMENTATION METHODS
public String getId() {
String msg = sm.getString("applicationSession.session.ise");
throw new IllegalStateException(msg);
return id;
} else {
return creationTime;
public long getCreationTime() {
return new SessionContextImpl();
public HttpSessionContext getSessionContext() {
public void invalidate() {
// remove everything in the session
serverSession.removeApplicationSession(context);
removeValue(name);
String name = (String)enum.nextElement();
Enumeration enum = values.keys();
while (enum.hasMoreElements()) {
public boolean isNew() {
valid = false;
if (! valid) {
return true;
return false;
if (thisAccessTime == creationTime) {
public void setAttribute(String name, Object value) {
setAttribute(name, value);
public void putValue(String name, Object value) {
removeValue(name); // remove any existing binding
throw new IllegalArgumentException(msg);
String msg = sm.getString("applicationSession.value.iae");
if (name == null) {
new HttpSessionBindingEvent(this, name);
((HttpSessionBindingListener)value).valueBound(e);
HttpSessionBindingEvent e =
if (value != null && value instanceof HttpSessionBindingListener) {
values.put(name, value);
public Object getAttribute(String name) {
return getAttribute(name);
public Object getValue(String name) {
return values.get(name);
public String[] getValueNames() {
names.addElement(e.nextElement());
while (e.hasMoreElements()) {
Vector names = new Vector();
Enumeration e = getAttributeNames();
names.copyInto(valueNames);
String[] valueNames = new String[names.size()];
public Enumeration getAttributeNames() {
return valueNames;
return (Enumeration)valuesClone.keys();
Hashtable valuesClone = (Hashtable)values.clone();
public void removeAttribute(String name) {
removeAttribute(name);
public void removeValue(String name) {
Object o = values.get(name);
((HttpSessionBindingListener)o).valueUnbound(e);
new HttpSessionBindingEvent(this,name);
if (o instanceof HttpSessionBindingListener) {
public void setMaxInactiveInterval(int interval) {
values.remove(name);
public int getMaxInactiveInterval() {
inactiveInterval = interval;
return inactiveInterval; //
ApplicationSession
import java.io.IOException;
import java.util.Vector;
import java.util.Hashtable;
import java.util.Enumeration;
import java.io.ObjectInputStream;
import java.io.ObjectOutputStream;
import java.io.Serializable;
import org.apache.tomcat.catalina.*;
import javax.servlet.http.HttpSessionContext;
import javax.servlet.http.HttpSessionBindingListener;
import javax.servlet.ServletException;
import javax.servlet.http.HttpSession;
import javax.servlet.http.HttpSessionBindingEvent;
* <b>IMPLEMENTATION NOTE</b>: An instance of this class represents both the
* <p>
* to a different JVM for distributable session support.
* Standard implementation of the <b>Session</b> interface. This object is
* serializable, so that it can be stored in persistent storage or transferred
* However, because the class itself is not declared public, Java logic outside
* internal (Session) and application level (HttpSession) view of the session.
* of the <code>org.apache.tomcat.session</code> package cannot cast an
Craig R. McClanahan
$Revision: 1.2 $ $Date: 2000/05/15 17:54:10 $
* HttpSession view of this instance back to a Session view.
// Constructors
implements HttpSession, Session {
final class StandardSession
manager The manager with which this Session is associated
* Construct a new Session associated with the specified Manager.
this.manager = manager;
super();
public StandardSession(Manager manager) {
// Instance Variables
private Hashtable attributes = new Hashtable();
* The collection of user data attributes associated with this Session.
* The time this session was created, in milliseconds since midnight,
private long creationTime = 0L;
* January 1, 1970 GMT.
private String id = null;
* The session identifier of this Session.
private static final String info = "StandardSession/1.0";
* Descriptive information describing this Session implementation.
* The last accessed time for this Session.
private Manager manager = null;
* The Manager with which this Session is associated.
* The maximum time interval, in seconds, between client requests before
private int maxInactiveInterval = -1;
* indicates that the session should never time out.
* the servlet container may invalidate this session. A negative time
private boolean isNew = true;
* Flag indicating whether this session is new or not.
private boolean isValid = false;
* Flag indicating whether this session is valid or not.
* The string manager for this package.
private static HttpSessionContext sessionContext = null;
* The HTTP session context associated with this session.
private long thisAccessedTime = creationTime;
* The current accessed time for this session.
// Session Properties
* Set the creation time for this session. This method is called by the
time The new creation time
* Manager when an existing Session instance is reused.
this.creationTime = time;
this.thisAccessedTime = time;
public void setCreationTime(long time) {
* Return the session identifier for this session.
return (this.id);
* Set the session identifier for this session.
id The new session identifier
((ManagerBase) manager).remove(this);
(manager instanceof ManagerBase))
if ((this.id != null) && (manager != null) &&
public void setId(String id) {
((ManagerBase) manager).add(this);
if ((manager != null) && (manager instanceof ManagerBase))
* the corresponding version number, in the format
* Return descriptive information about this Session implementation and
return (this.info);
public String getInfo() {
* <code><description>/<version></code>.
* Return the Manager within which this Session is valid.
return (this.manager);
public Manager getManager() {
public void setManager(Manager manager) {
manager The new Manager
* Set the Manager within which this Session is valid.
* Return the maximum time interval, in seconds, between client requests
IllegalStateException if this method is called on
* time indicates that the session should never time out.
* before the servlet container will invalidate the session. A negative
* an invalidated session
return (this.maxInactiveInterval);
* Set the maximum time interval, in seconds, between client requests
interval The new maximum interval
this.maxInactiveInterval = interval;
* Return the <code>HttpSession</code> for which this object
return ((HttpSession) this);
public HttpSession getSession() {
* is the facade.
// Session Public Methods
* without triggering an exception if the session has already expired.
* Perform the internal processing required to invalidate this session,
public void expire() {
// Remove this session from our manager's active sessions
String attr = (String) attrs.nextElement();
while (attrs.hasMoreElements()) {
Vector results = new Vector();
Enumeration attrs = getAttributeNames();
// Unbind any objects associated with this session
String name = (String) names.nextElement();
while (names.hasMoreElements()) {
Enumeration names = results.elements();
results.addElement(attr);
setValid(false);
// Mark this session as invalid
* preparation for reuse of this object.
* Release all object references, and initialize instance variables, in
public void recycle() {
manager = null;
maxInactiveInterval = -1;
id = null;
creationTime = 0L;
attributes.clear();
// Reset the instance variables associated with this Session
((ManagerBase) manager).recycle(this);
// Tell our Manager that this Session has been recycled
isValid = false;
isNew = true;
// Session Package Methods
* Return the <code>isValid</code> flag for this session.
return (this.isValid);
boolean isValid() {
void setNew(boolean isNew) {
isNew The new value for the <code>isNew</code> flag
* Set the <code>isNew</code> flag for this session.
this.isNew = isNew;
void setValid(boolean isValid) {
isValid The new value for the <code>isValid</code> flag
* Set the <code>isValid</code> flag for this session.
this.isValid = isValid;
// HttpSession Properties
IllegalStateException if this method is called on an
* midnight, January 1, 1970 GMT.
* Return the time when this session was created, in milliseconds since
* invalidated session
return (this.creationTime);
* replacement. It will be removed in a future version of the
As of Version 2.1, this method is deprecated and has no
* Return the session context with which this session is associated.
* Java Servlet API.
sessionContext = new StandardSessionContext();
return (sessionContext);
if (sessionContext == null)
// HttpSession Public Methods
* <code>null</code> if no object is bound with that name.
* Return the object bound with the specified name in this session, or
name Name of the attribute to be returned
return (attributes.get(name));
* containing the names of the objects bound to this session.
* Return an <code>Enumeration</code> of <code>String</code> objects
return (attributes.keys());
name Name of the value to be returned
* <code>getAttribute()</code>
As of Version 2.2, this method is replaced by
return (getAttribute(name));
* are no such objects, a zero-length array is returned.
* Return the set of names of objects bound to this session. If there
* <code>getAttributeNames()</code>
names[i] = (String) results.elementAt(i);
return (names);
for (int i = 0; i < names.length; i++)
String names[] = new String[results.size()];
* Invalidates this session and unbinds any objects bound to it.
expire();
// Cause this session to expire
* session, or if the client chooses not to join the session. For
* Return <code>true</code> if the client does not yet know about the
* request.
* has disabled the use of cookies, then a session would be new on each
* example, if the server used only cookie-based sessions, and the client
return (this.isNew);
* Bind an object to this session, using the specified name. If an object
* replaced.
* of the same name is already bound to this session, the object is
name Name to which the object is bound, cannot be null
value Object to be bound, cannot be null
* <code>valueBound()</code> on the object.
* After this method executes, and if the object implements
* <code>HttpSessionBindingListener</code>, the container calls
* <code>setAttribute()</code>
* does nothing.
* the session does not have an object bound with this name, this method
* Remove the object bound with the specified name from this session. If
name Name of the object to remove from this session.
* <code>valueUnbound()</code> on the object.
if (object == null)
Object object = attributes.get(name);
synchronized (attributes) {
(new HttpSessionBindingEvent((HttpSession) this, name));
((HttpSessionBindingListener) object).valueUnbound
if (object instanceof HttpSessionBindingListener) {
attributes.remove(name);
// System.out.println( "Removing attribute " + name );
return;
* <code>removeAttribute()</code>
IllegalArgumentException if an attempt is made to add a
* non-serializable object in an environment marked distributable.
throw new IllegalArgumentException
(sm.getString("standardSession.setAttribute.iae"));
!(value instanceof Serializable))
if ((manager != null) && manager.getDistributable() &&
if (value instanceof HttpSessionBindingListener)
((HttpSessionBindingListener) value).valueBound
attributes.put(name, value);
// HttpSession Private Methods
* <b>IMPLEMENTATION NOTE</b>: The reference to the owning Manager
* is not restored by this method, and must be set explicitly.
* object input stream.
* Read a serialized version of this session object from the specified
ClassNotFoundException if an unknown class is specified
stream The input stream to read from
// Deserialize the scalar instance variables (except Manager)
creationTime = ((Long) stream.readObject()).longValue();
throws ClassNotFoundException, IOException {
private void readObject(ObjectInputStream stream)
IOException if an input/output error occurs
isValid = ((Boolean) stream.readObject()).booleanValue();
id = (String) stream.readObject();
String name = (String) stream.readObject();
Object value = (Object) stream.readObject();
for (int i = 0; i < n; i++) {
int n = ((Integer) stream.readObject()).intValue();
// Deserialize the attribute count and attribute values
* object output stream.
* Write a serialized version of this session object to the specified
* <b>IMPLEMENTATION NOTE</b>: Any attribute that is not Serializable
* will be silently ignored. If you do not want any such attributes,
* explicitly.
* <code>readObject()</code>, you must set the associated Manager
* <b>IMPLEMENTATION NOTE</b>: The owning Manager will not be stored
* in the serialized representation of this Session. After calling
* Manager is set to <code>true</code>.
stream The output stream to write to
* be sure the <code>distributable</code> property of our associated
stream.writeObject(new Integer(maxInactiveInterval));
stream.writeObject(id);
stream.writeObject(new Long(creationTime));
private void writeObject(ObjectOutputStream stream) throws IOException {
// Write the scalar instance variables (except Manager)
stream.writeObject(new Boolean(isValid));
// Accumulate the names of serializable attributes
stream.writeObject(new Boolean(isNew));
if (value instanceof Serializable)
Object value = attributes.get(attr);
stream.writeObject(new Integer(results.size()));
// Serialize the attribute count and the attribute values
stream.writeObject(attributes.get(name));
stream.writeObject(name);
Enumeration getAttributeNames() |
String[] getValueNames() |
Object getAttribute(String) |
long getCreationTime() |
crosscut invalidate(StandardSession s): s & (int getMaxInactiveInterval() |
static advice(StandardSession s): invalidate(s) {
before {
void setAttribute(String, Object));
void removeAttribute(String) |
void invalidate() |
boolean isNew() |
+ ".ise"));
(s.sm.getString("standardSession."
+ thisJoinPoint.methodName
throw new IllegalStateException
if (!s.isValid())
// Private Class
* when <code>HttpSession.getSessionContext()</code> is called.
* This class is a dummy implementation of the <code>HttpSessionContext</code>
* interface, to conform to the requirement that such an object be returned
As of Java Servlet API 2.1 with no replacement. The
private Vector dummy = new Vector();
final class StandardSessionContext implements HttpSessionContext {
* interface will be removed in a future version of this API.
* and will be removed in a future version of the API.
* This method must return an empty <code>Enumeration</code>
* within this context.
* Return the session identifiers of all sessions defined
As of Java Servlet API 2.1 with no replacement.
return (dummy.elements());
public Enumeration getIds() {
* This method must return null and will be removed in a
id Session identifier for which to look up a session
* specified session identifier.
* Return the <code>HttpSession</code> associated with the
public HttpSession getSession(String id) {
* future version of the API.
return (null);
StandardSession
import javax.servlet.http.Cookie;
import org.w3c.dom.Node;
import org.w3c.dom.NamedNodeMap;
* an optional, configurable, maximum number of active sessions allowed.
* no session persistence or distributable capabilities, but does support
* Standard implementation of the <b>Manager</b> interface that provides
* checkInterval="60" maxActiveSessions="-1"
* maxInactiveInterval="-1" />
* <Manager className="org.apache.tomcat.session.StandardManager"
* <code>
* in the following format:
* Lifecycle configuration of this component assumes an XML node
* where you can adjust the following parameters, with default values
* </code>
* be active at once, or -1 for no limit. [-1]
* <li><b>maxInactiveInterval</b> - The default maximum number of seconds of
* inactivity before which the servlet container is allowed to time out
* <li><b>maxActiveSessions</b> - The maximum number of sessions allowed to
* thread checks for expired sessions. [60]
* <li><b>checkInterval</b> - The interval (in seconds) between background
* in square brackets:
* <ul>
* </ul>
* descriptor, if any. [-1]
* the default session timeout specified in the web application deployment
* a session, or -1 for no limit. This value should be overridden from
implements Lifecycle, Runnable {
extends ManagerBase
public final class StandardManager
$Revision: $ $Date: 2000/05/02 21:28:30 $
* The interval (in seconds) between checks for expired sessions.
private boolean configured = false;
* Has this component been configured yet?
private int checkInterval = 60;
private static final String info = "StandardManager/1.0";
* The descriptive information about this implementation.
protected int maxActiveSessions = -1;
* The maximum number of active Sessions allowed, or -1 for no limit.
private boolean started = false;
* Has this component been started yet?
private Thread thread = null;
* The background thread.
* The background thread completion semaphore.
private boolean threadDone = false;
* Name to register for the background thread.
// Properties
private String threadName = "StandardManager";
return (this.checkInterval);
public int getCheckInterval() {
* Return the check interval (in seconds) for this Manager.
checkInterval The new check interval
* Set the check interval (in seconds) for this Manager.
public void setCheckInterval(int checkInterval) {
this.checkInterval = checkInterval;
* Return descriptive information about this Manager implementation and
public int getMaxActiveSessions() {
* no limit.
* Return the maximum number of active Sessions allowed, or -1 for
return (this.maxActiveSessions);
* Set the maximum number of actives Sessions allowed, or -1 for
this.maxActiveSessions = max;
public void setMaxActiveSessions(int max) {
max The new maximum number of sessions
// Public Methods
* method of the returned session. If a new session cannot be created
* for any reason, return <code>null</code>.
* id will be assigned by this method, and available via the getId()
* settings specified by this Manager's properties. The session
* Construct and return a new session object, based on the default
if ((maxActiveSessions >= 0) &&
(sessions.size() >= maxActiveSessions))
public Session createSession() {
* instantiated for any reason
IllegalStateException if a new session cannot be
return (super.createSession());
(sm.getString("standardManager.createSession.ise"));
// Lifecycle Methods
* Configure this component, based on the specified configuration
IllegalStateException if this component has already been
* (<B>FIXME: What object type should this really be?)
parameters Configuration parameters for this component
* parameters. This method should be called immediately after the
* component instance is created, and before <code>start()</code>
* is called.
LifecycleException if this component detects a fatal error
* configured and/or started
if (configured)
// Validate and update our current component state
throws LifecycleException {
* in the configuration parameters it was given
public void configure(Node parameters)
if (!("Manager".equals(parameters.getNodeName())))
// Parse and process our configuration parameters
if (parameters == null)
(sm.getString("standardManager.alreadyConfigured"));
configured = true;
throw new LifecycleException
NamedNodeMap attributes = parameters.getAttributes();
} catch (Throwable t) {
; // XXX - Throw exception?
setCheckInterval(Integer.parseInt(node.getNodeValue()));
if (node != null) {
node = attributes.getNamedItem("checkInterval");
Node node = null;
try {
setMaxActiveSessions(Integer.parseInt(node.getNodeValue()));
node = attributes.getNamedItem("maxActiveSessions");
setMaxInactiveInterval(Integer.parseInt(node.getNodeValue()));
node = attributes.getNamedItem("maxInactiveInterval");
* component. This method should be called after <code>configure()</code>,
* Prepare for the beginning of active use of the public methods of this
* started
* configured (if required for this component)
IllegalStateException if this component has not yet been
* and before any of the public methods of the component are utilized.
public void start() throws LifecycleException {
* that prevents this component from being used
if (started)
(sm.getString("standardManager.alreadyStarted"));
(sm.getString("standardManager.notConfigured"));
if (!configured)
threadStart();
// Start the background reaper thread
started = true;
* instance of this component.
* Gracefully terminate the active use of the public methods of this
* component. This method should be the last one called on a given
* been stopped
IllegalStateException if this component has already
IllegalStateException if this component has not been started
if (!started)
public void stop() throws LifecycleException {
* that needs to be reported
started = false;
(sm.getString("standardManager.notStarted"));
for (int i = 0; i < sessions.length; i++) {
StandardSession session = (StandardSession) sessions[i];
Session sessions[] = findSessions();
// Expire all active sessions
threadStop();
// Stop the background reaper thread
continue;
session.expire();
if (!session.isValid())
// Private Methods
private void threadSleep() {
* Sleep for the duration specified by the <code>checkInterval</code>
* property. ;
} catch (InterruptedException e) {
Thread.sleep(checkInterval * 1000L);
private void threadStart() {
* session timeouts.
* Start the background thread that will periodically check for
thread.start();
thread.setDaemon(true);
threadDone = false;
if (thread != null)
thread = new Thread(this, threadName);
* Stop the background thread that is periodically checking for
threadDone = true;
if (thread == null)
private void threadStop() {
thread.join();
thread.interrupt();
thread = null;
// Background Thread
public void run() {
* The background thread that checks for session timeouts and shutdown.
processExpires();
threadSleep();
while (!threadDone) {
// Loop until the termination semaphore is set
StandardManager
StandardSessionManager
import org.apache.tomcat.util.SessionUtil;
import org.apache.tomcat.core.SessionManager;
import org.apache.tomcat.core.Response;
import org.apache.tomcat.core.Context;
import org.apache.tomcat.core.Request;
* XXX - At present, use of <code>StandardManager</code> is hard coded,
* that adapts to the new component-based Manager implementation.
* Specialized implementation of org.apache.tomcat.core.SessionManager
* and lifecycle configuration is not supported.
* the core level. The Tomcat.Next "Manager" interface acts more like a
* collection class, and has minimal knowledge of the detailed request
* paradigm, I would suggest moving the logic implemented here back into
* <b>IMPLEMENTATION NOTE</b>: Once we commit to the new Manager/Session
* a Context to tell the Manager that we create what the default session
* XXX - At present, there is no way (via the SessionManager interface) for
* processing semantics of handling sessions.
* should be.
* timeout for this web application (specified in the deployment descriptor)
implements SessionManager {
public final class StandardSessionManager
* Create a new SessionManager that adapts to the corresponding Manager
manager = new StandardManager();
public StandardSessionManager() {
* implementation.
if (manager instanceof Lifecycle) {
} catch (LifecycleException e) {
throw new IllegalStateException("" + e);
((Lifecycle) manager).start();
((Lifecycle) manager).configure(null);
* The Manager implementation we are actually using.
// XXX should we throw exception or just return null ??
public HttpSession findSession( Context ctx, String id ) {
} catch (IOException e) {
return session.getSession();
if(session!=null)
Session session = manager.findSession(id);
return manager.createSession().getSession();
public HttpSession createSession(Context ctx) {
* Remove all sessions because our associated Context is being shut down.
public void removeSessions(Context ctx) {
ctx The context that is being shut down
// The manager will still run after that ( i.e. keep database
// connection open
// contexts, we just want to remove the sessions of ctx!
// XXX XXX a manager may be shared by multiple
((Lifecycle) manager).stop();
* Used by context to configure the session manager's inactivity timeout.
* session manager according to this value.
* descriptor (web.xml). This method lets the Context conforgure the
* Context on the other hand has it's timeout set by the deployment
* The SessionManager may have some default session time out, the
if(-1 != minutes) {
// The manager works with seconds...
public void setSessionTimeOut(int minutes) {
minutes The session inactivity timeout in minutes.
manager.setMaxInactiveInterval(minutes * 60);
ServerSessionManager
import org.apache.tomcat.util.*;
public class ServerSessionManager implements SessionManager {
private static ServerSessionManager manager; // = new ServerSessionManager();
public static ServerSessionManager getManager() {
manager = new ServerSessionManager();
static {
protected int inactiveInterval = -1;
private Hashtable sessions = new Hashtable();
return manager;
reaper.setServerSessionManager(this);
reaper.start();
reaper = Reaper.getReaper();
private Reaper reaper;
private ServerSessionManager() {
String sessionId = SessionIdGenerator.generateId();
session.setMaxInactiveInterval(inactiveInterval);
return session.getApplicationSession( ctx, true );
if(-1 != inactiveInterval) {
sessions.put(sessionId, session);
ServerSession session = new ServerSession(sessionId);
if(sSession==null) return null;
public HttpSession findSession(Context ctx, String id) {
return sSession.getApplicationSession(ctx, false);
ServerSession sSession=(ServerSession)sessions.get(id);
// XXX
Enumeration enum = sessions.keys();
synchronized void reap() {
// solution for this, but we'll determine something else later.
// sync'd for safty -- no other thread should be getting something
// from this while we are reaping. This isn't the most optimal
session.reap();
Object key = enum.nextElement();
session.validate();
ServerSession session = (ServerSession)sessions.get(key);
String id = session.getId();
synchronized void removeSession(ServerSession session) {
sessions.remove(id);
public void removeSessions(Context context) {
session.invalidate();
appSession.invalidate();
if (appSession != null) {
session.getApplicationSession(context, false);
ApplicationSession appSession =
inactiveInterval = (minutes * 60);
SessionInterceptor
package org.apache.tomcat.request;
* It also marks the session as accessed.
* in the Request.
* Will process the request and determine the session Id, and set it
// GS, separates the session id from the jvm route
static final char SESSIONID_ROUTE_SEP = '.';
public class SessionInterceptor extends BaseInterceptor implements RequestInterceptor {
* add new interceptors for other methods.
* This implementation only handles Cookies sessions, please extend or
public SessionInterceptor() {
int debug=0;
ContextManager cm;
public void setContextManager( ContextManager cm ) {
debug=i;
System.out.println("Set debug to " + i);
public void setDebug( int i ) {
String sessionId = null;
public int requestMap(Request request ) {
this.cm=cm;
if (cookie.getName().equals("JSESSIONID")) {
sessionId = cookie.getValue();
Cookie cookie = cookies[i];
for( int i=0; i<cookies.length; i++ ) {
Cookie cookies[]=request.getCookies(); // assert !=null
request.setRequestedSessionIdFromCookie(true);
if (sessionId!=null){
sessionId=validateSessionId(request, sessionId);
sessionId=request.getRequestURI().substring(foundAt+sig.length());
request.setRequestURI(request.getRequestURI().substring(0, foundAt));
// rewrite URL, do I need to do anything more?
if ((foundAt=request.getRequestURI().indexOf(sig))!=-1){
if( debug>0 ) cm.log(" XXX RURI=" + request.getRequestURI());
String sig=";jsessionid=";
int foundAt=-1;
return 0;
request.setRequestedSessionIdFromURL(true);
sessionId, or null if not valid
* It will also clean up the session from load-balancing strings.
/** Validate and fix the session id. If the session is not valid return null.
// We may still set it and just return session invalid.
// XXX what is the correct behavior if the session is invalid ?
// GS, We piggyback the JVM id on top of the session cookie
private String validateSessionId(Request request, String sessionId){
if(idex > 0) {
sessionId = sessionId.substring(0, idex);
int idex = sessionId.lastIndexOf(SESSIONID_ROUTE_SEP);
if (null != sessionId) {
// Separate them ...
if( debug>0 ) cm.log(" Orig sessionId " + sessionId );
if (sessionId != null && sessionId.length()!=0) {
Context ctx=request.getContext();
SessionManager sM = ctx.getSessionManager();
if(null != sM.findSession(ctx, sessionId)) {
// cookie. We must check for validity in the current context.
// context and one for the real context... or old session
// GS, We are in a problem here, we may actually get
// multiple Session cookies (one for the root
return sessionId;
request.setRequestedSessionId(sessionId);
if( debug>0 ) cm.log(" Final session id " + sessionId );
return null;
public int beforeBody( Request rrequest, Response response ) {
if( debug>0 ) cm.log("Before Body " + reqSessionId );
if( reqSessionId==null)
String reqSessionId = response.getSessionId();
if(sessionPath.length() == 0) {
sessionPath = "/";
String sessionPath = rrequest.getContext().getPath();
// multiple session cookies can be used, one for each
// context.
// GS, set the path attribute to the cookie. This way
if(null != jvmRoute) {
// GS, piggyback the jvm route on the session id.
reqSessionId = reqSessionId + SESSIONID_ROUTE_SEP + jvmRoute;
String jvmRoute = rrequest.getJvmRoute();
if(!sessionPath.equals("/")) {
cookie.setMaxAge(-1);
reqSessionId);
Cookie cookie = new Cookie("JSESSIONID",
CookieTools.getCookieHeaderValue(cookie));
cookie.setVersion(0);
response.addHeader( CookieTools.getCookieHeaderName(cookie),
cookie.setVersion(1);
cookie.setPath(sessionPath); {
if( ctx.getDebug() > 0 ) ctx.log("Removing sessions from " + ctx );
public void contextShutdown( Context ctx )
throws TomcatException
/** Notification of context shutdown
ctx.getSessionManager().removeSessions(ctx);
ServerSession
* Core implementation of a server session
public class ServerSession {
private Hashtable appSessions = new Hashtable();
ServerSession(String id) {
private int inactiveInterval = -1;
(ApplicationSession)appSessions.get(context);
boolean create) {
public ApplicationSession getApplicationSession(Context context,
appSessions.put(context, appSession);
appSession = new ApplicationSession(id, this, context);
// sync to ensure valid?
if (appSession == null && create) {
return appSession;
// a new appSession
// inactive interval -- if so, invalidate and create
// make sure that we haven't gone over the end of our
appSessions.remove(context);
void removeApplicationSession(Context context) {
void validate()
synchronized void invalidate() {
Enumeration enum = appSessions.keys();
(ApplicationSession)appSessions.get(key);
String msg = sm.getString("serverSession.value.iae");
public Enumeration getValueNames() {
return values.keys();
appSession.validate();
private long lastAccessed = creationTime;
validate();
thisAccessTime = System.currentTimeMillis();
lastAccessed = thisAccessTime;
void accessed() {
// set last accessed to thisAccessTime as it will be left over
// from the previous access
int thisInterval =
if (inactiveInterval != -1) {
// if we have an inactive interval, check to see if we've exceeded it
void validate() {
if (thisInterval > inactiveInterval) {
invalidate();
(int)(System.currentTimeMillis() - lastAccessed) / 1000;
return lastAccessed;
public long getLastAccessedTime() {
// we've exceeded it
// if we have an inactive interval, check to see if
ServerSessionManager ssm =
ssm.removeSession(this);
ServerSessionManager.getManager();
private long lastAccessedTime = creationTime;
* GMT. Actions that your application takes, such as getting or setting
* session, as the number of milliseconds since midnight, January 1, 1970
* Return the last time the client sent a request associated with this
* a value associated with the session, do not affect the access time.
return (this.lastAccessedTime);
this.lastAccessedTime = time;
* session, even if the application does not reference it.
* Update the accessed time information for this session. This method
* should be called by the context when a request comes in for a particular
public void access() {
lastAccessedTime = 0L;
this.isNew=false;
this.thisAccessedTime = System.currentTimeMillis();
this.lastAccessedTime = this.thisAccessedTime;
isNew = ((Boolean) stream.readObject()).booleanValue();
lastAccessedTime = ((Long) stream.readObject()).longValue();
maxInactiveInterval = ((Integer) stream.readObject()).intValue();
stream.writeObject(new Long(lastAccessedTime));
ApplicationSession apS=(ApplicationSession)findSession( ctx, id);
public void accessed( Context ctx, Request req, String id ) {
sM.accessed(ctx, request, sessionId );
// cache it - no need to compute it again
apS.accessed();
servS.accessed();
if( apS==null) return;
ServerSession servS=apS.getServerSession();
req.setSession( apS );
long timeNow = System.currentTimeMillis();
* Invalidate all sessions that have expired.
private void processExpires() {
int timeIdle = // Truncate, do not round up
int maxInactiveInterval = session.getMaxInactiveInterval();
if (maxInactiveInterval < 0)
if (timeIdle >= maxInactiveInterval)
(int) ((timeNow - session.getLastAccessedTime()) / 1000L);
* called for each request by a RequestInterceptor.
* Mark the specified session's last accessed time. This should be
HttpSession session=findSession(ctx, id);
if( session == null) return;
public void accessed(Context ctx, Request req, String id) {
session The session to be marked
// cache the HttpSession - avoid another find
((Session) session).access();
if (session instanceof Session)
req.setSession( session ); /* *
* ====================================================================
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* reserved.
* Copyright (c) 1999 The Apache Software Foundation. All rights
* The Apache Software License, Version 1.1
* 1. Redistributions of source code must retain the above copyright
* are met:
* the documentation and/or other materials provided with the
* distribution.
* notice, this list of conditions and the following disclaimer in
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer.
itself,
* if and wherever such third-party acknowlegements normally appear.
* Apache Software Foundation (
* Alternately, this acknowlegement may appear in the software
* "This product includes software developed by the
* 3. The end-user documentation included with the redistribution, if
* any, must include the following acknowlegement:
* Foundation" must not be used to endorse or promote products
* 4. The names "The Jakarta Project", "Tomcat", and "Apache Software
* nor may "Apache" appear in their names without prior written
* permission of the Apache Group.
* 5. Products derived from this software may not be called "Apache"
* permission, please contact
derived
* from this software without prior written permission. For written
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
* ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
* THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESSED OR IMPLIED
* DISCLAIMED. IN NO EVENT SHALL THE APACHE SOFTWARE FOUNDATION OR
* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* This software consists of voluntary contributions made by many
* SUCH DAMAGE.
* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
* USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
* ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
* information on the Apache Software Foundation, please see
* individuals on behalf of the Apache Software Foundation. For more
* [Additional notices, if required by prior licensing conditions] */
* <
package org.apache.tomcat.session;
import java.net.*;
import java.util.*;
import java.io.*;
import org.apache.tomcat.util.StringManager;
import org.apache.tomcat.core.*;
* Core implementation of an application level session
James Duncan Davidson
/**
import javax.servlet.http.*;
import javax.servlet.*;
James Todd
Jason Hunter
private String id;
private Hashtable values = new Hashtable();
private StringManager sm =
public class ApplicationSession implements HttpSession {
StringManager.getManager("org.apache.tomcat.session");
private ServerSession serverSession;
private boolean valid = true;
private int inactiveInterval = -1;
private long lastAccessed = creationTime;
private Context context;
private long creationTime = System.currentTimeMillis();;
private long thisAccessTime = creationTime;
this.inactiveInterval = context.getSessionTimeOut();
this.context = context;
this.id = id;
Context context) {
ApplicationSession(String id, ServerSession serverSession,
this.serverSession = serverSession; }
this.inactiveInterval *= 60;
if (this.inactiveInterval != -1) {
return serverSession;
ServerSession getServerSession() {
* inactivities can be dealt with accordingly.
* Called by context when request comes in so that accesses and
void accessed() {
thisAccessTime = System.currentTimeMillis();
lastAccessed = thisAccessTime;
// set last accessed to thisAccessTime as it will be left over
// from the previous access
int thisInterval =
if (inactiveInterval != -1) {
void validate() {
validate();
invalidate();
if (thisInterval > inactiveInterval) {
(int)(System.currentTimeMillis() - lastAccessed) / 1000;
// if we have an inactive interval, check to see if we've exceeded it
if (valid) {
// HTTP SESSION IMPLEMENTATION METHODS
String msg = sm.getString("applicationSession.session.ise");
public String getId() {
throw new IllegalStateException(msg);
} else {
return id;
return creationTime;
public long getCreationTime() {
public long getLastAccessedTime() {
return new SessionContextImpl();
public HttpSessionContext getSessionContext() {
return lastAccessed;
public void invalidate() {
// remove everything in the session
serverSession.removeApplicationSession(context);
removeValue(name);
String name = (String)enum.nextElement();
Enumeration enum = values.keys();
while (enum.hasMoreElements()) {
public boolean isNew() {
valid = false;
if (! valid) {
return true;
return false;
if (thisAccessTime == creationTime) {
public void setAttribute(String name, Object value) {
setAttribute(name, value);
public void putValue(String name, Object value) {
removeValue(name); // remove any existing binding
throw new IllegalArgumentException(msg);
String msg = sm.getString("applicationSession.value.iae");
if (name == null) {
new HttpSessionBindingEvent(this, name);
((HttpSessionBindingListener)value).valueBound(e);
HttpSessionBindingEvent e =
if (value != null && value instanceof HttpSessionBindingListener) {
values.put(name, value);
public Object getAttribute(String name) {
return getAttribute(name);
public Object getValue(String name) {
return values.get(name);
public String[] getValueNames() {
names.addElement(e.nextElement());
while (e.hasMoreElements()) {
Vector names = new Vector();
Enumeration e = getAttributeNames();
names.copyInto(valueNames);
String[] valueNames = new String[names.size()];
public Enumeration getAttributeNames() {
return valueNames;
return (Enumeration)valuesClone.keys();
Hashtable valuesClone = (Hashtable)values.clone();
public void removeAttribute(String name) {
removeAttribute(name);
public void removeValue(String name) {
Object o = values.get(name);
((HttpSessionBindingListener)o).valueUnbound(e);
new HttpSessionBindingEvent(this,name);
if (o instanceof HttpSessionBindingListener) {
public void setMaxInactiveInterval(int interval) {
values.remove(name);
public int getMaxInactiveInterval() {
inactiveInterval = interval;
return inactiveInterval; //
ApplicationSession
import java.io.IOException;
import java.util.Vector;
import javax.servlet.ServletException;
import java.util.Hashtable;
import java.util.Enumeration;
import java.io.ObjectInputStream;
import java.io.ObjectOutputStream;
import java.io.Serializable;
import javax.servlet.http.HttpSession;
import javax.servlet.http.HttpSessionContext;
import org.apache.tomcat.catalina.*;
import javax.servlet.http.HttpSessionBindingListener;
import javax.servlet.http.HttpSessionBindingEvent;
* serializable, so that it can be stored in persistent storage or transferred
* Standard implementation of the <b>Session</b> interface. This object is
* internal (Session) and application level (HttpSession) view of the session.
* <b>IMPLEMENTATION NOTE</b>: An instance of this class represents both the
* <p>
* to a different JVM for distributable session support.
* HttpSession view of this instance back to a Session view.
* of the <code>org.apache.tomcat.session</code> package cannot cast an
* However, because the class itself is not declared public, Java logic outside
$Revision: 1.2 $ $Date: 2000/05/15 17:54:10 $
Craig R. McClanahan
implements HttpSession, Session {
final class StandardSession
// Constructors
manager The manager with which this Session is associated
* Construct a new Session associated with the specified Manager.
this.manager = manager;
super();
public StandardSession(Manager manager) {
// Instance Variables
private Hashtable attributes = new Hashtable();
* The collection of user data attributes associated with this Session.
* The time this session was created, in milliseconds since midnight,
private long creationTime = 0L;
* January 1, 1970 GMT.
private String id = null;
* The session identifier of this Session.
private static final String info = "StandardSession/1.0";
* Descriptive information describing this Session implementation.
private long lastAccessedTime = creationTime;
* The last accessed time for this Session.
private Manager manager = null;
* The Manager with which this Session is associated.
* the servlet container may invalidate this session. A negative time
* The maximum time interval, in seconds, between client requests before
* indicates that the session should never time out.
private boolean isNew = true;
* Flag indicating whether this session is new or not.
private int maxInactiveInterval = -1;
private boolean isValid = false;
* Flag indicating whether this session is valid or not.
* The string manager for this package.
private static HttpSessionContext sessionContext = null;
* The HTTP session context associated with this session.
private long thisAccessedTime = creationTime;
* The current accessed time for this session.
// Session Properties
* Set the creation time for this session. This method is called by the
public void setCreationTime(long time) {
time The new creation time
* Manager when an existing Session instance is reused.
this.thisAccessedTime = time;
this.lastAccessedTime = time;
this.creationTime = time;
* Return the session identifier for this session.
return (this.id);
(manager instanceof ManagerBase))
if ((this.id != null) && (manager != null) &&
public void setId(String id) {
* Set the session identifier for this session.
id The new session identifier
((ManagerBase) manager).remove(this);
((ManagerBase) manager).add(this);
if ((manager != null) && (manager instanceof ManagerBase))
* <code><description>/<version></code>.
* Return descriptive information about this Session implementation and
* the corresponding version number, in the format
return (this.info);
public String getInfo() {
* a value associated with the session, do not affect the access time.
* GMT. Actions that your application takes, such as getting or setting
* session, as the number of milliseconds since midnight, January 1, 1970
* Return the last time the client sent a request associated with this
return (this.lastAccessedTime);
return (this.manager);
public Manager getManager() {
* Return the Manager within which this Session is valid.
public void setManager(Manager manager) {
manager The new Manager
* Set the Manager within which this Session is valid.
* Return the maximum time interval, in seconds, between client requests
IllegalStateException if this method is called on
* time indicates that the session should never time out.
* before the servlet container will invalidate the session. A negative
return (this.maxInactiveInterval);
* an invalidated session
* Set the maximum time interval, in seconds, between client requests
interval The new maximum interval
this.maxInactiveInterval = interval;
public HttpSession getSession() {
* is the facade.
* Return the <code>HttpSession</code> for which this object
return ((HttpSession) this);
// Session Public Methods
* session, even if the application does not reference it.
* should be called by the context when a request comes in for a particular
* Update the accessed time information for this session. This method
this.isNew=false;
this.thisAccessedTime = System.currentTimeMillis();
this.lastAccessedTime = this.thisAccessedTime;
public void access() {
* without triggering an exception if the session has already expired.
* Perform the internal processing required to invalidate this session,
// Remove this session from our manager's active sessions
public void expire() {
while (attrs.hasMoreElements()) {
Enumeration attrs = getAttributeNames();
Vector results = new Vector();
// Unbind any objects associated with this session
while (names.hasMoreElements()) {
String name = (String) names.nextElement();
Enumeration names = results.elements();
results.addElement(attr);
String attr = (String) attrs.nextElement();
setValid(false);
// Mark this session as invalid
public void recycle() {
* preparation for reuse of this object.
* Release all object references, and initialize instance variables, in
// Reset the instance variables associated with this Session
manager = null;
isNew = true;
maxInactiveInterval = -1;
lastAccessedTime = 0L;
attributes.clear();
creationTime = 0L;
id = null;
isValid = false;
((ManagerBase) manager).recycle(this);
// Tell our Manager that this Session has been recycled
// Session Package Methods
boolean isValid() {
* Return the <code>isValid</code> flag for this session.
return (this.isValid);
void setNew(boolean isNew) {
isNew The new value for the <code>isNew</code> flag
* Set the <code>isNew</code> flag for this session.
this.isNew = isNew;
isValid The new value for the <code>isValid</code> flag
* Set the <code>isValid</code> flag for this session.
this.isValid = isValid;
void setValid(boolean isValid) {
// HttpSession Properties
IllegalStateException if this method is called on an
* Return the time when this session was created, in milliseconds since
* midnight, January 1, 1970 GMT.
* invalidated session
return (this.creationTime);
* Return the session context with which this session is associated.
* Java Servlet API.
* replacement. It will be removed in a future version of the
As of Version 2.1, this method is deprecated and has no
return (sessionContext);
sessionContext = new StandardSessionContext();
if (sessionContext == null)
// HttpSession Public Methods
* <code>null</code> if no object is bound with that name.
* Return the object bound with the specified name in this session, or
name Name of the attribute to be returned
return (attributes.get(name));
* Return an <code>Enumeration</code> of <code>String</code> objects
* containing the names of the objects bound to this session.
return (attributes.keys());
name Name of the value to be returned
* <code>getAttribute()</code>
As of Version 2.2, this method is replaced by
return (getAttribute(name));
* are no such objects, a zero-length array is returned.
* Return the set of names of objects bound to this session. If there
* <code>getAttributeNames()</code>
names[i] = (String) results.elementAt(i);
return (names);
for (int i = 0; i < names.length; i++)
String names[] = new String[results.size()];
* Invalidates this session and unbinds any objects bound to it.
expire();
// Cause this session to expire
* session, or if the client chooses not to join the session. For
* Return <code>true</code> if the client does not yet know about the
* request.
* has disabled the use of cookies, then a session would be new on each
* example, if the server used only cookie-based sessions, and the client
return (this.isNew);
* Bind an object to this session, using the specified name. If an object
* replaced.
* of the same name is already bound to this session, the object is
name Name to which the object is bound, cannot be null
value Object to be bound, cannot be null
* <code>valueBound()</code> on the object.
* After this method executes, and if the object implements
* <code>HttpSessionBindingListener</code>, the container calls
* <code>setAttribute()</code>
* does nothing.
* the session does not have an object bound with this name, this method
* Remove the object bound with the specified name from this session. If
name Name of the object to remove from this session.
* <code>valueUnbound()</code> on the object.
Object object = attributes.get(name);
synchronized (attributes) {
((HttpSessionBindingListener) object).valueUnbound
(new HttpSessionBindingEvent((HttpSession) this, name));
if (object instanceof HttpSessionBindingListener) {
// System.out.println( "Removing attribute " + name );
return;
attributes.remove(name);
if (object == null)
* <code>removeAttribute()</code>
IllegalArgumentException if an attempt is made to add a
* non-serializable object in an environment marked distributable.
throw new IllegalArgumentException
(sm.getString("standardSession.setAttribute.iae"));
!(value instanceof Serializable))
if ((manager != null) && manager.getDistributable() &&
((HttpSessionBindingListener) value).valueBound
if (value instanceof HttpSessionBindingListener)
attributes.put(name, value);
// HttpSession Private Methods
* is not restored by this method, and must be set explicitly.
* <b>IMPLEMENTATION NOTE</b>: The reference to the owning Manager
* object input stream.
* Read a serialized version of this session object from the specified
ClassNotFoundException if an unknown class is specified
IOException if an input/output error occurs
stream The input stream to read from
id = (String) stream.readObject();
lastAccessedTime = ((Long) stream.readObject()).longValue();
creationTime = ((Long) stream.readObject()).longValue();
// Deserialize the scalar instance variables (except Manager)
private void readObject(ObjectInputStream stream)
throws ClassNotFoundException, IOException {
isNew = ((Boolean) stream.readObject()).booleanValue();
maxInactiveInterval = ((Integer) stream.readObject()).intValue();
String name = (String) stream.readObject();
Object value = (Object) stream.readObject();
for (int i = 0; i < n; i++) {
int n = ((Integer) stream.readObject()).intValue();
// Deserialize the attribute count and attribute values
isValid = ((Boolean) stream.readObject()).booleanValue();
* object output stream.
* Write a serialized version of this session object to the specified
* explicitly.
* <b>IMPLEMENTATION NOTE</b>: Any attribute that is not Serializable
* <code>readObject()</code>, you must set the associated Manager
* in the serialized representation of this Session. After calling
* <b>IMPLEMENTATION NOTE</b>: The owning Manager will not be stored
stream The output stream to write to
* Manager is set to <code>true</code>.
* be sure the <code>distributable</code> property of our associated
* will be silently ignored. If you do not want any such attributes,
stream.writeObject(id);
stream.writeObject(new Long(creationTime));
private void writeObject(ObjectOutputStream stream) throws IOException {
stream.writeObject(new Boolean(isValid));
stream.writeObject(new Integer(maxInactiveInterval));
stream.writeObject(new Long(lastAccessedTime));
// Write the scalar instance variables (except Manager)
// Accumulate the names of serializable attributes
stream.writeObject(new Boolean(isNew));
if (value instanceof Serializable)
Object value = attributes.get(attr);
stream.writeObject(new Integer(results.size()));
// Serialize the attribute count and the attribute values
stream.writeObject(attributes.get(name));
stream.writeObject(name);
Enumeration getAttributeNames() |
Object getAttribute(String) |
crosscut invalidate(StandardSession s): s & (int getMaxInactiveInterval() |
long getCreationTime() |
String[] getValueNames() |
static advice(StandardSession s): invalidate(s) {
if (!s.isValid())
before {
void setAttribute(String, Object));
void removeAttribute(String) |
void invalidate() |
boolean isNew() |
+ ".ise"));
+ thisJoinPoint.methodName
throw new IllegalStateException
(s.sm.getString("standardSession."
// Private Class
* This class is a dummy implementation of the <code>HttpSessionContext</code>
* when <code>HttpSession.getSessionContext()</code> is called.
* interface, to conform to the requirement that such an object be returned
* interface will be removed in a future version of this API.
As of Java Servlet API 2.1 with no replacement. The
* Return the session identifiers of all sessions defined
* within this context.
private Vector dummy = new Vector();
final class StandardSessionContext implements HttpSessionContext {
public Enumeration getIds() {
* and will be removed in a future version of the API.
As of Java Servlet API 2.1 with no replacement.
* This method must return an empty <code>Enumeration</code>
return (dummy.elements());
id Session identifier for which to look up a session
* Return the <code>HttpSession</code> associated with the
* specified session identifier.
return (null);
public HttpSession getSession(String id) {
* future version of the API.
* This method must return null and will be removed in a
StandardSession
import javax.servlet.http.Cookie;
import org.w3c.dom.Node;
import org.w3c.dom.NamedNodeMap;
* an optional, configurable, maximum number of active sessions allowed.
* no session persistence or distributable capabilities, but does support
* Standard implementation of the <b>Manager</b> interface that provides
* checkInterval="60" maxActiveSessions="-1"
* maxInactiveInterval="-1" />
* <Manager className="org.apache.tomcat.session.StandardManager"
* <code>
* in the following format:
* Lifecycle configuration of this component assumes an XML node
* where you can adjust the following parameters, with default values
* </code>
* be active at once, or -1 for no limit. [-1]
* <li><b>maxInactiveInterval</b> - The default maximum number of seconds of
* inactivity before which the servlet container is allowed to time out
* <li><b>maxActiveSessions</b> - The maximum number of sessions allowed to
* thread checks for expired sessions. [60]
* <li><b>checkInterval</b> - The interval (in seconds) between background
* in square brackets:
* <ul>
* </ul>
* descriptor, if any. [-1]
* the default session timeout specified in the web application deployment
* a session, or -1 for no limit. This value should be overridden from
implements Lifecycle, Runnable {
extends ManagerBase
public final class StandardManager
$Revision: $ $Date: 2000/05/02 21:28:30 $
* The interval (in seconds) between checks for expired sessions.
private boolean configured = false;
* Has this component been configured yet?
private int checkInterval = 60;
private static final String info = "StandardManager/1.0";
* The descriptive information about this implementation.
protected int maxActiveSessions = -1;
* The maximum number of active Sessions allowed, or -1 for no limit.
private boolean started = false;
* Has this component been started yet?
private Thread thread = null;
* The background thread.
* The background thread completion semaphore.
private boolean threadDone = false;
* Name to register for the background thread.
// Properties
private String threadName = "StandardManager";
return (this.checkInterval);
public int getCheckInterval() {
* Return the check interval (in seconds) for this Manager.
checkInterval The new check interval
* Set the check interval (in seconds) for this Manager.
public void setCheckInterval(int checkInterval) {
this.checkInterval = checkInterval;
* Return descriptive information about this Manager implementation and
public int getMaxActiveSessions() {
* no limit.
* Return the maximum number of active Sessions allowed, or -1 for
return (this.maxActiveSessions);
* Set the maximum number of actives Sessions allowed, or -1 for
this.maxActiveSessions = max;
public void setMaxActiveSessions(int max) {
max The new maximum number of sessions
// Public Methods
* method of the returned session. If a new session cannot be created
* for any reason, return <code>null</code>.
* id will be assigned by this method, and available via the getId()
* settings specified by this Manager's properties. The session
* Construct and return a new session object, based on the default
if ((maxActiveSessions >= 0) &&
(sessions.size() >= maxActiveSessions))
public Session createSession() {
* instantiated for any reason
IllegalStateException if a new session cannot be
return (super.createSession());
(sm.getString("standardManager.createSession.ise"));
// Lifecycle Methods
* Configure this component, based on the specified configuration
IllegalStateException if this component has already been
* (<B>FIXME: What object type should this really be?)
parameters Configuration parameters for this component
* parameters. This method should be called immediately after the
* component instance is created, and before <code>start()</code>
* is called.
LifecycleException if this component detects a fatal error
* configured and/or started
if (configured)
// Validate and update our current component state
throws LifecycleException {
* in the configuration parameters it was given
public void configure(Node parameters)
configured = true;
(sm.getString("standardManager.alreadyConfigured"));
throw new LifecycleException
NamedNodeMap attributes = parameters.getAttributes();
if (!("Manager".equals(parameters.getNodeName())))
// Parse and process our configuration parameters
if (parameters == null)
Node node = null;
} catch (Throwable t) {
; // XXX - Throw exception?
setCheckInterval(Integer.parseInt(node.getNodeValue()));
try {
if (node != null) {
node = attributes.getNamedItem("checkInterval");
setMaxActiveSessions(Integer.parseInt(node.getNodeValue()));
node = attributes.getNamedItem("maxActiveSessions");
setMaxInactiveInterval(Integer.parseInt(node.getNodeValue()));
node = attributes.getNamedItem("maxInactiveInterval");
IllegalStateException if this component has not yet been
* configured (if required for this component)
* and before any of the public methods of the component are utilized.
* component. This method should be called after <code>configure()</code>,
* Prepare for the beginning of active use of the public methods of this
public void start() throws LifecycleException {
* that prevents this component from being used
* started
(sm.getString("standardManager.notConfigured"));
if (!configured)
// Start the background reaper thread
threadStart();
started = true;
(sm.getString("standardManager.alreadyStarted"));
if (started)
* component. This method should be the last one called on a given
* Gracefully terminate the active use of the public methods of this
* been stopped
* that needs to be reported
IllegalStateException if this component has already
* instance of this component.
IllegalStateException if this component has not been started
public void stop() throws LifecycleException {
// Stop the background reaper thread
threadStop();
started = false;
if (!started)
(sm.getString("standardManager.notStarted"));
// Expire all active sessions
continue;
session.expire();
if (!session.isValid())
StandardSession session = (StandardSession) sessions[i];
Session sessions[] = findSessions();
for (int i = 0; i < sessions.length; i++) {
// Private Methods
* Invalidate all sessions that have expired.
long timeNow = System.currentTimeMillis();
private void processExpires() {
(int) ((timeNow - session.getLastAccessedTime()) / 1000L);
if (timeIdle >= maxInactiveInterval)
int timeIdle = // Truncate, do not round up
if (maxInactiveInterval < 0)
int maxInactiveInterval = session.getMaxInactiveInterval();
private void threadSleep() {
* property.
* Sleep for the duration specified by the <code>checkInterval</code> ;
} catch (InterruptedException e) {
Thread.sleep(checkInterval * 1000L);
if (thread != null)
private void threadStart() {
* Start the background thread that will periodically check for
* session timeouts.
thread.start();
thread.setDaemon(true);
thread = new Thread(this, threadName);
threadDone = false;
private void threadStop() {
* Stop the background thread that is periodically checking for
if (thread == null)
thread.join();
thread.interrupt();
threadDone = true;
thread = null;
// Background Thread
* The background thread that checks for session timeouts and shutdown.
processExpires();
threadSleep();
// Loop until the termination semaphore is set
public void run() {
while (!threadDone) {
StandardManager
StandardSessionManager
import org.apache.tomcat.util.SessionUtil;
import org.apache.tomcat.core.SessionManager;
import org.apache.tomcat.core.Response;
import org.apache.tomcat.core.Context;
import org.apache.tomcat.core.Request;
* XXX - At present, use of <code>StandardManager</code> is hard coded,
* that adapts to the new component-based Manager implementation.
* Specialized implementation of org.apache.tomcat.core.SessionManager
* and lifecycle configuration is not supported.
* the core level. The Tomcat.Next "Manager" interface acts more like a
* collection class, and has minimal knowledge of the detailed request
* paradigm, I would suggest moving the logic implemented here back into
* <b>IMPLEMENTATION NOTE</b>: Once we commit to the new Manager/Session
* a Context to tell the Manager that we create what the default session
* XXX - At present, there is no way (via the SessionManager interface) for
* processing semantics of handling sessions.
* should be.
* timeout for this web application (specified in the deployment descriptor)
implements SessionManager {
public final class StandardSessionManager
* Create a new SessionManager that adapts to the corresponding Manager
manager = new StandardManager();
public StandardSessionManager() {
* implementation.
if (manager instanceof Lifecycle) {
} catch (LifecycleException e) {
throw new IllegalStateException("" + e);
((Lifecycle) manager).start();
((Lifecycle) manager).configure(null);
* The Manager implementation we are actually using.
* Mark the specified session's last accessed time. This should be
public void accessed(Context ctx, Request req, String id) {
session The session to be marked
* called for each request by a RequestInterceptor.
HttpSession session=findSession(ctx, id);
// cache the HttpSession - avoid another find
((Session) session).access();
if (session instanceof Session)
if( session == null) return;
req.setSession( session );
Session session = manager.findSession(id);
public HttpSession findSession( Context ctx, String id ) {
// XXX should we throw exception or just return null ??
} catch (IOException e) {
if(session!=null)
return session.getSession();
return manager.createSession().getSession();
public HttpSession createSession(Context ctx) {
* Remove all sessions because our associated Context is being shut down.
ctx The context that is being shut down
// contexts, we just want to remove the sessions of ctx!
// The manager will still run after that ( i.e. keep database
// XXX XXX a manager may be shared by multiple
public void removeSessions(Context ctx) {
// connection open
((Lifecycle) manager).stop();
* Used by context to configure the session manager's inactivity timeout.
* descriptor (web.xml). This method lets the Context conforgure the
* Context on the other hand has it's timeout set by the deployment
* The SessionManager may have some default session time out, the
public void setSessionTimeOut(int minutes) {
minutes The session inactivity timeout in minutes.
* session manager according to this value.
manager.setMaxInactiveInterval(minutes * 60);
if(-1 != minutes) {
// The manager works with seconds...
ServerSessionManager
import org.apache.tomcat.util.*;
public class ServerSessionManager implements SessionManager {
manager = new ServerSessionManager();
static {
protected int inactiveInterval = -1;
private static ServerSessionManager manager; // = new ServerSessionManager();
private Reaper reaper;
private Hashtable sessions = new Hashtable();
return manager;
public static ServerSessionManager getManager() {
reaper.start();
reaper.setServerSessionManager(this);
private ServerSessionManager() {
reaper = Reaper.getReaper();
if( apS==null) return;
ServerSession servS=apS.getServerSession();
ApplicationSession apS=(ApplicationSession)findSession( ctx, id);
public void accessed( Context ctx, Request req, String id ) {
req.setSession( apS );
// cache it - no need to compute it again
apS.accessed();
servS.accessed();
if(-1 != inactiveInterval) {
session.setMaxInactiveInterval(inactiveInterval);
sessions.put(sessionId, session);
ServerSession session = new ServerSession(sessionId);
String sessionId = SessionIdGenerator.generateId();
if(sSession==null) return null;
return sSession.getApplicationSession(ctx, false);
public HttpSession findSession(Context ctx, String id) {
return session.getApplicationSession( ctx, true );
ServerSession sSession=(ServerSession)sessions.get(id);
// XXX
Enumeration enum = sessions.keys();
synchronized void reap() {
// solution for this, but we'll determine something else later.
// sync'd for safty -- no other thread should be getting something
// from this while we are reaping. This isn't the most optimal
session.reap();
Object key = enum.nextElement();
session.validate();
ServerSession session = (ServerSession)sessions.get(key);
String id = session.getId();
synchronized void removeSession(ServerSession session) {
sessions.remove(id);
public void removeSessions(Context context) {
session.invalidate();
appSession.invalidate();
if (appSession != null) {
session.getApplicationSession(context, false);
ApplicationSession appSession =
inactiveInterval = (minutes * 60);
SessionInterceptor
package org.apache.tomcat.request;
* It also marks the session as accessed.
* in the Request.
* Will process the request and determine the session Id, and set it
// GS, separates the session id from the jvm route
static final char SESSIONID_ROUTE_SEP = '.';
public class SessionInterceptor extends BaseInterceptor implements RequestInterceptor {
* add new interceptors for other methods.
* This implementation only handles Cookies sessions, please extend or
public SessionInterceptor() {
int debug=0;
ContextManager cm;
public void setContextManager( ContextManager cm ) {
debug=i;
System.out.println("Set debug to " + i);
public void setDebug( int i ) {
String sessionId = null;
public int requestMap(Request request ) {
this.cm=cm;
if (cookie.getName().equals("JSESSIONID")) {
sessionId = cookie.getValue();
Cookie cookie = cookies[i];
for( int i=0; i<cookies.length; i++ ) {
Cookie cookies[]=request.getCookies(); // assert !=null
request.setRequestedSessionIdFromCookie(true);
if (sessionId!=null){
sessionId=validateSessionId(request, sessionId);
sessionId=request.getRequestURI().substring(foundAt+sig.length());
request.setRequestURI(request.getRequestURI().substring(0, foundAt));
// rewrite URL, do I need to do anything more?
if ((foundAt=request.getRequestURI().indexOf(sig))!=-1){
if( debug>0 ) cm.log(" XXX RURI=" + request.getRequestURI());
String sig=";jsessionid=";
int foundAt=-1;
return 0;
request.setRequestedSessionIdFromURL(true);
* It will also clean up the session from load-balancing strings.
sessionId, or null if not valid
/** Validate and fix the session id. If the session is not valid return null.
// We may still set it and just return session invalid.
// XXX what is the correct behavior if the session is invalid ?
// Separate them ...
// GS, We piggyback the JVM id on top of the session cookie
private String validateSessionId(Request request, String sessionId){
if(idex > 0) {
sessionId = sessionId.substring(0, idex);
int idex = sessionId.lastIndexOf(SESSIONID_ROUTE_SEP);
if( debug>0 ) cm.log(" Orig sessionId " + sessionId );
if (null != sessionId) {
if (sessionId != null && sessionId.length()!=0) {
SessionManager sM = ctx.getSessionManager();
if(null != sM.findSession(ctx, sessionId)) {
sM.accessed(ctx, request, sessionId );
Context ctx=request.getContext();
// cookie. We must check for validity in the current context.
// multiple Session cookies (one for the root
// GS, We are in a problem here, we may actually get
// context and one for the real context... or old session
return sessionId;
if( debug>0 ) cm.log(" Final session id " + sessionId );
request.setRequestedSessionId(sessionId);
return null;
public int beforeBody( Request rrequest, Response response ) {
if( debug>0 ) cm.log("Before Body " + reqSessionId );
if( reqSessionId==null)
String reqSessionId = response.getSessionId();
if(sessionPath.length() == 0) {
sessionPath = "/";
String sessionPath = rrequest.getContext().getPath();
// multiple session cookies can be used, one for each
// context.
// GS, set the path attribute to the cookie. This way
if(null != jvmRoute) {
// GS, piggyback the jvm route on the session id.
reqSessionId = reqSessionId + SESSIONID_ROUTE_SEP + jvmRoute;
String jvmRoute = rrequest.getJvmRoute();
if(!sessionPath.equals("/")) {
cookie.setMaxAge(-1);
reqSessionId);
Cookie cookie = new Cookie("JSESSIONID",
CookieTools.getCookieHeaderValue(cookie));
cookie.setVersion(0);
response.addHeader( CookieTools.getCookieHeaderName(cookie),
cookie.setVersion(1);
cookie.setPath(sessionPath); {
if( ctx.getDebug() > 0 ) ctx.log("Removing sessions from " + ctx );
public void contextShutdown( Context ctx )
throws TomcatException
/** Notification of context shutdown
ctx.getSessionManager().removeSessions(ctx);
ServerSession
* Core implementation of a server session
public class ServerSession {
private Hashtable appSessions = new Hashtable();
ServerSession(String id) {
(ApplicationSession)appSessions.get(context);
public ApplicationSession getApplicationSession(Context context,
boolean create) {
appSession = new ApplicationSession(id, this, context);
appSessions.put(context, appSession);
// sync to ensure valid?
if (appSession == null && create) {
return appSession;
// a new appSession
// inactive interval -- if so, invalidate and create
// make sure that we haven't gone over the end of our
appSessions.remove(context);
void removeApplicationSession(Context context) {
void validate()
// if we have an inactive interval, check to see if
// we've exceeded it
ssm.removeSession(this);
ServerSessionManager ssm =
ServerSessionManager.getManager();
Enumeration enum = appSessions.keys();
synchronized void invalidate() {
(ApplicationSession)appSessions.get(key);
String msg = sm.getString("serverSession.value.iae");
return values.keys();
public Enumeration getValueNames() {
appSession.validate();
what you are going to see is that if you set out to do this, it is possible to do so. AspectJ is a mix of old and new ideas – its good that some of them are old – in a new package that explicitly supports crosscutting modularity.
AspectJ Tutorial

10. modular aspects HTTPRequest SessionInterceptor getCookies()
getRequestURI()(doc)
getSession() getRequestedSessionId()
...
requestMap(request)
beforeBody(req, resp)
...
Session
HTTPResponse
getAttribute(name)
setAttribute(name, val)
invalidate()
...
getRequest()
setContentType(contentType)
getOutptutStream() setSessionId(id)
...
Servlet
AspectJ Tutorial

11. AspectJ™ is… a small and well-integrated extension to Java
a general-purpose AO language
just as Java is a general-purpose OO language
freely available implementation
compiler is Open Source
includes IDE support
emacs, JBuilder 3.5, Forte 4J
user feedback is driving language design
currently at 0.7b8 release
1.0 planned for June 2001
AspectJ Tutorial

12. expected benefits of using AOP
good modularity, even for crosscutting concerns
less tangled code
more natural code
shorter code
easier maintenance and evolution
easier to reason about, debug, change
more reusable
library aspects
plug and play aspects when appropriate
AspectJ Tutorial

13. outline I AOP overview II AspectJ language mechanisms
AspectJ Tutorial
outline
I AOP overview
brief motivation, essence of AOP idea
II AspectJ language mechanisms
basic concepts, language semantics
III development environment
IDE support, running the compiler, debugging etc.
IV using aspects
aspect examples, how to use AspectJ to program aspects, exercises to solidify the ideas
V related work
survey of other activities in the AOP community
AspectJ Tutorial

14. looking ahead problem structure AspectJ mechanisms Part IV:
crosscutting in the design, and how to use AspectJ to capture that
AspectJ mechanisms
Part II:
crosscutting in the code mechanisms AspectJ provides
AspectJ Tutorial

15. Basic Mechanisms of AspectJ
Part II
Basic Mechanisms of AspectJ
AspectJ Tutorial

16. goals of this chapter present basic language mechanisms
using one simple example
emphasis on what the mechanisms do
small scale motivation
later chapters elaborate on
environment, tools
larger examples, design and SE issues
AspectJ Tutorial

17. basic mechanisms 1 overlay onto Java 4 small additions to Java
join points
“points in the execution” of Java programs
4 small additions to Java
pointcuts
primitive pointcuts
pick out sets of join points and values at those points
user-defined pointcuts
named collections of join points and values
advice
additional action to take at join points in a pointcut
introduction
additional fields/methods/constructors for classes
aspect
a crosscutting type
comprised of advice, introduction,
field,constructor and method declarations
AspectJ Tutorial

18. display must be updated when objects move
a simple figure editor
class Line implements FigureElement{
private Point _p1, _p2;
Point getP1() { return _p1; }
Point getP2() { return _p2; }
void setP1(Point p1) { _p1 = p1; }
void setP2(Point p2) { _p2 = p2; } }
class Point implements FigureElement {
private int _x = 0, _y = 0;
int getX() { return _x; }
int getY() { return _y; }
void setX(int x) { _x = x; }
void setY(int y) { _y = y; }
display must be updated when objects move
AspectJ Tutorial

19. move tracking … collection of figure elements other examples
that change periodically
must monitor changes to refresh the display as needed
collection can be complex
hierarchical
asynchronous events
other examples
session liveness
value caching …
AspectJ Tutorial

20. join points key points in dynamic call graph a Figure a Line
AspectJ Tutorial
join points
key points in dynamic call graph
object receives a method call
a Figure
object receives a method call
and returns or throws
dispatch
and returns or throws
a method call to another object
a Line
billions and billions
dispatch
and a return or exception is received by this method
AspectJ Tutorial

21. join point terminology
key points in dynamic call graph
a Line
dispatch
method call reception join points
method call join points
9 kinds of join points
method & constructor call reception join points
method & constructor call join points
method & constructor execution join points
field get & set join points
exception handler execution join points
AspectJ Tutorial

22. join point terminology
AspectJ Tutorial
join point terminology
key points in dynamic call graph
all join points on this slide are within the control flow of this join point
a Line
a Point
a Figure
the method call join point corresponding to this method reception join point
a Point
repeated calls to the same method on the same object result in multiple join points
billions and billions
AspectJ Tutorial

23. the pointcut construct
names certain join points
each time a Line receives a “void setP1(Point)” or “void setP2(Point)” method call
pointcut moves():
receptions(void Line.setP1(Point)) ||
receptions(void Line.setP2(Point));
name and parameters
reception of a “void Line.setP1(Point)” call or
reception of a “void Line.setP2(Point)” call
AspectJ Tutorial

24. pointcut designators user-defined pointcut designator
pointcut moves():
receptions(void Line.setP1(Point)) ||
receptions(void Line.setP2(Point));
user-defined pointcut designator
primitive pointcut designator, can also be:
- calls, executions - instanceof,
- gets, sets - within, withincode
- handlers - cflow, cflowtop
AspectJ Tutorial

25. after advice runs “on the way back out”
action to take after computation under join points
after advice runs “on the way back out”
a Line
pointcut moves():
receptions(void Line.setP1(Point)) ||
receptions(void Line.setP2(Point));
static after(): moves() {
<runs after moves> }
AspectJ Tutorial

26. a simple aspect MoveTracking v1
aspect defines a special class that can crosscut other classes
aspect MoveTracking {
private static boolean _flag = false;
public static boolean testAndClear() {
boolean result = _flag;
_flag = false;
return result; }
pointcut moves():
receptions(void Line.setP1(Point)) ||
receptions(void Line.setP2(Point));
static after(): moves() {
_flag = true;
box means complete running code
AspectJ Tutorial

27. without AspectJ what you would expect MoveTracking v1
class Line {
private Point _p1, _p2;
Point getP1() { return _p1; }
Point getP2() { return _p2; }
void setP1(Point p1) {
_p1 = p1;
MoveTracking.setFlag(); }
void setP2(Point p2) {
_p2 = p2;
class MoveTracking {
private static boolean _flag = false;
public static void setFlag() {
_flag = true; }
public static boolean testAndClear() { boolean result = _flag;
_flag = false;
return result;
what you would expect
calls to set flag are tangled through the code
“what is going on” is less explicit
AspectJ Tutorial

28. the pointcut construct
can cut across multiple classes
pointcut moves():
receptions(void Line.setP1(Point)) ||
receptions(void Line.setP2(Point)) ||
receptions(void Point.setX(int)) ||
receptions(void Point.setY(int));
AspectJ Tutorial

29. a multi-class aspect MoveTracking v2 aspect MoveTracking {
private static boolean _flag = false;
public static boolean testAndClear() {
boolean result = _flag;
_flag = false;
return result; }
pointcut moves():
receptions(void Line.setP1(Point)) ||
receptions(void Line.setP2(Point)) ||
receptions(void Point.setX(int)) ||
receptions(void Point.setY(int));
static after(): moves() {
_flag = true;
AspectJ Tutorial

30. using context in advice
demonstrate first, explain in detail afterwards
pointcut can explicitly expose certain values
advice can use value
pointcut moves(FigureElement figElt):
instanceof(figElt) &&
(receptions(void Line.setP1(Point)) ||
receptions(void Line.setP2(Point)) ||
receptions(void Point.setX(int)) ||
receptions(void Point.setY(int)));
static after(FigureElement fe): moves(fe) {
<fe is bound to the figure element> }
parameter mechanism is being used
AspectJ Tutorial

31. context & multiple classes
MoveTracking v3
aspect MoveTracking {
private static Set _movees = new HashSet();
public static Set getMovees() {
Set result = _movees;
_movees = new HashSet();
return result; }
pointcut moves(FigureElement figElt):
instanceof(figElt) &&
(receptions(void Line.setP1(Point)) ||
receptions(void Line.setP2(Point)) ||
receptions(void Point.setX(int)) ||
receptions(void Point.setY(int)));
static after(FigureElement fe): moves(fe) {
_movees.add(fe);
AspectJ Tutorial

32. parameters… pulls corresponding value out of join points
of user-defined pointcut designator
variable bound in user-defined pointcut designator
variable in place of type name in pointcut designator
pulls corresponding value out of join points
makes value accessible on pointcut
pointcut moves(Line l):
receptions(void l.setP1(Point)) ||
receptions(void l.setP2(Point));
static after(Line line): moves(line) { <line is bound to the line> }
pointcut parameters
variable in place of type name
AspectJ Tutorial

33. typed variable in place of type name
parameters…
of advice
variable bound in advice
variable in place of type name in pointcut designator
pulls corresponding value out of join points
makes value accessible within advice
pointcut moves(Line l):
receptions(void l.setP1(Point)) ||
receptions(void l.setP2(Point));
static after(Line line): moves(line) { <line is bound to the line> }
typed variable in place of type name
advice parameters
AspectJ Tutorial

34. parameters… value is ‘pulled’
right to left across ‘:’ left side : right side
from pointcut designators to user-defined pointcut designators
from pointcut to advice
pointcut moves(Line l):
receptions(void l.setP1(Point)) ||
receptions(void l.setP2(Point));
static after(Line line): moves(line) { <line is bound to the line> }
AspectJ Tutorial

35. instanceof primitive pointcut designator instanceof(<type name>)
any join point at which
currently executing object is ‘instanceof’ type (or class) name
instanceof(Point)
instanceof(Line)
instanceof(FigureElement)
“any join point” means it matches join points of all 9 kinds
method & constructor call join points
method & constructor call reception join points
method & constructor execution join points
field get & set join points
exception handler execution join points
AspectJ Tutorial

36. an idiom for… getting object in a polymorphic pointcut
instanceof(<supertype name>) &&
does not further restrict the join points
does pick up the currently executing object (this)
pointcut moves(FigureElement figElt):
instanceof(figElt) &&
(receptions(void Line.setP1(Point)) ||
receptions(void Line.setP2(Point)) ||
receptions(void Point.setX(int)) ||
receptions(void Point.setY(int)));
static after(FigureElement fe): moves(fe) {
<fe is bound to the figure element> }
AspectJ Tutorial

37. context & multiple classes
MoveTracking v3
aspect MoveTracking {
private static Set _movees = new HashSet();
public static Set getMovees() {
Set result = _movees;
_movees = new HashSet();
return result; }
pointcut moves(FigureElement figElt):
instanceof(figElt) &&
(receptions(void Line.setP1(Point)) ||
receptions(void Line.setP2(Point)) ||
receptions(void Point.setX(int)) ||
receptions(void Point.setY(int)));
static after(FigureElement fe): moves(fe) {
_movees.add(fe);
AspectJ Tutorial

38. without AspectJ class Line { class Point { private int _x = 0, _y = 0;
private Point _p1, _p2;
Point getP1() { return _p1; }
Point getP2() { return _p2; }
void setP1(Point p1) {
_p1 = p1; }
void setP2(Point p2) {
_p2 = p2;
class Point {
private int _x = 0, _y = 0;
int getX() { return _x; }
int getY() { return _y; }
void setX(int x) {
_x = x;
void setY(int y) {
_y = y;
AspectJ Tutorial

39. without AspectJ MoveTracking v1 class Line { class Point {
private Point _p1, _p2;
Point getP1() { return _p1; }
Point getP2() { return _p2; }
void setP1(Point p1) {
_p1 = p1;
MoveTracking.setFlag(); }
void setP2(Point p2) {
_p2 = p2;
class Point {
private int _x = 0, _y = 0;
int getX() { return _x; }
int getY() { return _y; }
void setX(int x) {
_x = x;
void setY(int y) {
_y = y;
class MoveTracking {
private static boolean _flag = false;
public static void setFlag() {
_flag = true; }
public static boolean testAndClear() { boolean result = _flag;
_flag = false;
return result;
AspectJ Tutorial

40. without AspectJ MoveTracking v2 class Line { class Point {
private Point _p1, _p2;
Point getP1() { return _p1; }
Point getP2() { return _p2; }
void setP1(Point p1) {
_p1 = p1;
MoveTracking.setFlag(); }
void setP2(Point p2) {
_p2 = p2;
class Point {
private int _x = 0, _y = 0;
int getX() { return _x; }
int getY() { return _y; }
void setX(int x) {
_x = x;
void setY(int y) {
_y = y;
class MoveTracking {
private static boolean _flag = false;
public static void setFlag() {
_flag = true; }
public static boolean testAndClear() { boolean result = _flag;
_flag = false;
return result;
AspectJ Tutorial

41. without AspectJ evolution is cumbersome MoveTracking v3
class Line {
private Point _p1, _p2;
Point getP1() { return _p1; }
Point getP2() { return _p2; }
void setP1(Point p1) {
_p1 = p1;
MoveTracking.collectOne(this); }
void setP2(Point p2) {
_p2 = p2;
class Point {
private int _x = 0, _y = 0;
int getX() { return _x; }
int getY() { return _y; }
void setX(int x) {
_x = x;
void setY(int y) {
_y = y;
class MoveTracking {
private static Set _movees = new HashSet();
public static void collectOne(Object o) {
_movees.add(o); }
public static Set getmovees() {
Set result = _movees;
_movees = new HashSet();
return result;
evolution is cumbersome
changes in all three classes
have to track all callers
change method name
add argument
AspectJ Tutorial

42. with AspectJ class Line { class Point { private int _x = 0, _y = 0;
private Point _p1, _p2;
Point getP1() { return _p1; }
Point getP2() { return _p2; }
void setP1(Point p1) {
_p1 = p1; }
void setP2(Point p2) {
_p2 = p2;
class Point {
private int _x = 0, _y = 0;
int getX() { return _x; }
int getY() { return _y; }
void setX(int x) {
_x = x;
void setY(int y) {
_y = y;
AspectJ Tutorial

43. with AspectJ MoveTracking v1 class Line { class Point {
private Point _p1, _p2;
Point getP1() { return _p1; }
Point getP2() { return _p2; }
void setP1(Point p1) {
_p1 = p1; }
void setP2(Point p2) {
_p2 = p2;
class Point {
private int _x = 0, _y = 0;
int getX() { return _x; }
int getY() { return _y; }
void setX(int x) {
_x = x;
void setY(int y) {
_y = y;
aspect MoveTracking {
private static boolean _flag = false;
public static boolean testAndClear() {
boolean result = _flag;
_flag = false;
return result; }
pointcut moves():
receptions(void Line.setP1(Point)) ||
receptions(void Line.setP2(Point));
static after(): moves() {
_flag = true;
AspectJ Tutorial

44. with AspectJ MoveTracking v2 class Line { class Point {
private Point _p1, _p2;
Point getP1() { return _p1; }
Point getP2() { return _p2; }
void setP1(Point p1) {
_p1 = p1; }
void setP2(Point p2) {
_p2 = p2;
class Point {
private int _x = 0, _y = 0;
int getX() { return _x; }
int getY() { return _y; }
void setX(int x) {
_x = x;
void setY(int y) {
_y = y;
aspect MoveTracking {
private static boolean _flag = false;
public static boolean testAndClear() {
boolean result = _flag;
_flag = false;
return result; }
pointcut moves():
receptions(void Line.setP1(Point)) ||
receptions(void Line.setP2(Point)) ||
receptions(void Point.setX(int)) ||
receptions(void Point.setY(int));
static after(): moves() {
_flag = true;
AspectJ Tutorial

45. with AspectJ evolution is more modular MoveTracking v3
class Line {
private Point _p1, _p2;
Point getP1() { return _p1; }
Point getP2() { return _p2; }
void setP1(Point p1) {
_p1 = p1; }
void setP2(Point p2) {
_p2 = p2;
class Point {
private int _x = 0, _y = 0;
int getX() { return _x; }
int getY() { return _y; }
void setX(int x) {
_x = x;
void setY(int y) {
_y = y;
aspect MoveTracking {
private static Set _movees = new HashSet();
public static Set getmovees() {
Set result = _movees;
_movees = new HashSet();
return result; }
pointcut moves(FigureElement figElt):
instanceof(figElt) &&
(receptions(void Line.setP1(Point)) ||
receptions(void Line.setP2(Point)) ||
receptions(void Point.setX(int)) ||
receptions(void Point.setY(int)));
static after(FigureElement fe): moves(fe) {
_movees.add(fe);
evolution is more modular
all changes in single aspect
AspectJ Tutorial

46. advice is before before proceeding at join point
additional action to take at join points
before before proceeding at join point
after returning a value to join point
after throwing a throwable to join point
after returning to join point either way
around on arrival at join point gets explicit control over when&if program proceeds
AspectJ Tutorial

47. contract checking pre-conditions post-conditions condition enforcement
simple example of before/after/around
pre-conditions
check whether parameter is valid
post-conditions
check whether values were set
condition enforcement
force parameters to be valid
AspectJ Tutorial

48. pre-condition using before advice
AspectJ Tutorial
pre-condition
using before advice
aspect PointBoundsPreCondition {
static before(Point p, int newX):
receptions(void p.setX(newX)) {
assert(newX >= MIN_X);
assert(newX <= MAX_X); }
static before(Point p, int newY):
receptions(void p.setY(newY)) {
assert(newY >= MIN_Y);
assert(newY <= MAX_Y);
static void assert(boolean v) {
if ( !v )
throw new RuntimeException();
what follows the ‘:’ is always a pointcut – primitive or user-defined
note this is the first place we are using anonymouse pointcuts
AspectJ Tutorial

49. post-condition using after advice aspect PointBoundsPostCondition {
static after(Point p, int newX):
receptions(void p.setX(newX)) {
assert(p.getX() == newX); }
static after(Point p, int newY):
receptions(void p.setY(newY)) {
assert(p.getY() == newY);
static void assert(boolean v) {
if ( !v )
throw new RuntimeException();
AspectJ Tutorial

50. condition enforcement
using around advice
aspect PointBoundsEnforcement {
static around(Point p, int newX) returns void:
receptions(void p.setX(newX)) {
thisJoinPoint.runNext(p, clip(newX, MIN_X, MAX_X)); }
static around(Point p, int newY) returns void:
receptions(void p.setY(newY)) {
thisJoinPoint.runNext(p, clip(newY, MIN_Y, MAX_Y));
static int clip(int val, int min, int max) {
return Math.max(min, Math.min(max, val));
AspectJ Tutorial

51. special static method <result type> runNext(arg1, arg2…)
AspectJ Tutorial
special static method
<result type> runNext(arg1, arg2…)
available only in around advice
means “run what would have run if this around advice had not been defined”
AspectJ Tutorial

52. other primitive pointcuts
AspectJ Tutorial
other primitive pointcuts
instanceof(<type or class name>)
within(<class name>) withincode(<method/constructor signature>)
any join point at which
currently executing object is ‘instanceof’ type or class name
currently executing code is contained within class name currently executing code is specified method or constructor
gets(int Point.x)
sets(int Point.x)
gets(int Point.x)[val]
sets(int Point.x)[oldVal][newVal]
field reference or assignment join points
Preview of what’s to come wrt sites. How sites are designated.
AspectJ Tutorial

53. using field set pointcuts
aspect PointCoordinateTracing {
pointcut coordChanges(Point p, int oldVal, int newVal):
sets(int p._x)[oldVal][newVal] ||
sets(int p._y)[oldVal][newVal];
static after(Point p, int oldVal, int newVal):
coordChanges(p, oldVal, newVal) {
System.out.println(“The ” + tjp.fieldName +
“ field of ” + p +
“ was changed from ” + oldVal +
“ to ” + newVal + “.”); }
AspectJ Tutorial

54. special value reflective* access to the join point thisJoinPoint.
AspectJ Tutorial
special value
reflective* access to the join point
thisJoinPoint.
String className
String methodName
String[] parameterNames
Class[] parameterTypes
Object[] parameters
...
available in any advice
thisJoinPoint is abbreviated to ‘tjp’ occasionally in these slides
* introspective subset of reflection consistent with Java
these will soon change to getClassName()…
AspectJ Tutorial

55. other primitive pointcuts
AspectJ Tutorial
other primitive pointcuts
calls(void Point.setX(int))
method/constructor call join points (at call site)
receptions(void Point.setX(int))
method/constructor call reception join points (at called object)
executions(void Point.setX(int))
method/constructor execution join points (at actual called method)
Preview of what’s to come wrt sites. How sites are designated.
AspectJ Tutorial

56. context sensitive aspects
MoveTracking v4a
aspect MoveTracking {
List _movers = new LinkedList();
List _movees = new LinkedList();
// …
pointcut moveCalls(Object mover, FigureElement movee):
instanceof(mover) &&
(lineMoveCalls(movee) || pointMoveCalls(movee));
pointcut lineMoveCalls(Line ln):
calls(void ln.setP1(Point)) || calls(void ln.setP2(Point));
pointcut pointMoveCalls(Point pt):
calls(void pt.setX(int)) || calls(void pt.setY(int));
static after(Object mover, FigureElement movee):
moveCalls(mover, movee) {
_movers.add(mover);
_movees.add(movee); }
AspectJ Tutorial

57. context sensitive aspects
MoveTracking v4b
aspect MoveTracking {
List _movers = new LinkedList();
List _movees = new LinkedList();
// …
pointcut moveCalls(Object mover, FigureElement movee):
instanceof(mover) &&
(calls(void ((Line)movee).setP1(Point)) ||
calls(void ((Line)movee).setP2(Point)) ||
calls(void ((Point)movee).setX(int)) ||
calls(void ((Point)movee).setY(int)));
static after(Object mover, FigureElement movee):
moveCalls(mover, movee) {
_movers.add(mover);
_movees.add(movee); }
does this make sense ?
AspectJ Tutorial

58. fine-grained protection
class Point implement FigureElement {
private int _x = 0, _y = 0;
int getX() { return _x; }
int getY() { return _y; }
void setX(int nv) { primitiveSetX(nv); }
void setY(int nv) { primitiveSetY(nv); }
void primitiveSetX(int x) { _x = x; }
void primitiveSetY(int y) { _y = y; } }
aspect PrimitiveSetterEnforcement {
pointcut illegalSets(Point pt):
!(withincode(void Point.primitiveSetX(int)) ||
withincode(void Point.primitiveSetY(int))) &&
(sets(int pt._x) || sets(int pt._y));
static before(Point p): illegalSets(p) {
throw new Error("Illegal primitive setter call.");
AspectJ Tutorial

59. other primitive pointcuts
AspectJ Tutorial
other primitive pointcuts
cflow(pointcut designator)
cflowtop(pointcut designator)
all join points within the dynamic control flow of any join point in pointcut designator
cflowtop doesn’t “start a new one” on re-entry
Preview of what’s to come wrt sites. How sites are designated.
AspectJ Tutorial

60. context sensitive aspects
MoveTracking v5
aspect MoveTracking {
private static Set _movees = new HashSet();
public static Set getMovees() {
Set result = _movees;
_movees = new HashSet();
return result; }
pointcut moves(FigureElement figElt):
instanceof(figElt) &&
(receptions(void Line.setP1(Point)) ||
receptions(void Line.setP2(Point)) ||
receptions(void Point.setX(int)) ||
receptions(void Point.setY(int)));
pointcut topLevelMoves(FigureElement figElt):
moves(figElt) && !cflow(moves(FigureElement));
static after(FigureElement fe): topLevelMoves(fe) {
_movees.add(fe);
AspectJ Tutorial

61. wildcarding in pointcuts
AspectJ Tutorial
wildcarding in pointcuts
“*” is wild card
“..” is multi-part wild card
instanceof(Point)
instanceof(graphics.geom.Point)
instanceof(graphics.geom.*) any type in graphics.geom
instanceof(graphics..*) any type in any sub-package
of graphics
receptions(void Point.setX(int))
receptions(public * Point.*(..)) any public method on Point
receptions(public * *..*.*(..)) any public method on any type
receptions(void Point.getX())
receptions(void Point.getY())
receptions(void Point.get*())
receptions(void get*()) any getter
receptions(Point.new(int, int))
receptions(new(..)) any constructor
Preview of what’s to come wrt sites. How sites are designated.
AspectJ Tutorial

62. property-based crosscutting
package com.xerox.scan;
public class C2 { …
public int frotz() {
A.doSomething(…); }
public int bar() {
package com.xerox.print;
public class C1 { …
public void foo() {
A.doSomething(…); }
package com.xerox.copy;
public class C3 { …
public String s1() {
A.doSomething(…); }
crosscuts of methods with a common property
public/private, return a certain value, in a particular package
logging, debugging, profiling
log on entry to every public method
AspectJ Tutorial

63. property-based crosscutting
AspectJ Tutorial
property-based crosscutting
aspect PublicErrorLogging {
static Log log = new Log();
pointcut publicInterface ():
receptions(public * com.xerox..*.*(..));
static after() throwing (Error e): publicInterface() {
log.write(e); }
neatly captures public interface of mypackage
maintenance is always where better modularity pays off best
consider code maintenance
another programmer adds a public method
i.e. extends public interface – this code will still work
another programmer reads this code
“what’s really going on” is explicit
AspectJ Tutorial

64. aspect state what if you want a per-object log?
aspect PublicErrorLogging
of eachobject(PublicErrorLogging.publicInterface()) {
Log log = new Log();
pointcut publicInterface ():
receptions(public * com.xerox..*.*(..));
after() throwing (Error e): publicInterface() {
log.write(e); }
one instance of the aspect for each object that ever executes at these points
body of advice is like body of a method: - “this” is bound to instance of aspect, - aspect fields are accessed freely
variable & advice are not static
AspectJ Tutorial

65. looking up aspect instances:
static Log getLog(Object obj) {
return (PublicErrorLogging.aspectOf(obj)).log; }
static method of aspects that are
of eachobject
of eachclass
of eachcflowroot
returns aspect instance or null
AspectJ Tutorial

66. of each relations eachobject(<pointcut>)
AspectJ Tutorial
of each relations
eachobject(<pointcut>)
one aspect instance for each
object that is ever “this” at
the join points
eachclass(<pointcut>)
one aspect instance for each class of
object that is ever “this” at the join points
eachcflowroot(<pointcut>)
one aspect instance for each join point in pointcut, is available at all joinpoints in <pointcut> && cflow(<pointcut>) p1 l1 l2 p3 p2
FigureElement
Point
Line
Preview of what’s to come wrt sites. How sites are designated.
AspectJ Tutorial

67. inheritance & specialization
pointcuts can have additional advice
aspect with
concrete pointcut
perhaps no advice on the pointcut
in figure editor
moves() can have advice from multiple aspects
module can expose certain well-defined pointcuts
abstract pointcuts can be specialized
abstract pointcut
concrete advice on the abstract pointcut
AspectJ Tutorial

68. a shared pointcut public class FigureEditor {
AspectJ Tutorial
a shared pointcut
public class FigureEditor {
public pointcut moves(FigureElement figElt):
instanceof(figElt) &&
(receptions(void Line.setP1(Point)) ||
receptions(void Line.setP2(Point)) ||
receptions(void Point.setX(int)) ||
receptions(void Point.setY(int)));
... }
aspect MoveTracking {
static after(FigureElement fe):
FigureEditor.moves(fe) { ... }
AspectJ Tutorial

69. a reusable aspect abstract
abstract public aspect RemoteExceptionLogging {
abstract pointcut logPoints();
static after() throwing (RemoteException e): logPoints() {
log.println(“Remote call failed in: ” thisJoinPoint.toString() “(” + e + “).”); }
abstract
public aspect MyRMILogging extends RemoteExceptionLogging {
pointcut logPoints():
receptions(* RegistryServer.*.*(..)) ||
receptions(private * RMIMessageBrokerImpl.*.*(..)); }
AspectJ Tutorial

70. public and private are with respect to enclosing aspect declaration
introduction
(like “open classes”) MoveTracking v6
aspect MoveTracking {
private static Set _movees = new HashSet();
public static Set getMovees() {
Set result = _movees;
_movees = new HashSet();
return result; }
introduction FigureElement {
private Object lastMovedBy;
public Object getLastMovedBy() { return lastMovedBy; }
pointcut MoveCalls(Object mover, FigureElement movee):
instanceof(mover) &&
(lineMoveCalls(movee) || pointMoveCalls(movee));
pointcut lineMoveCalls(Line ln):
calls(void ln.setP1(Point)) || calls(void ln.setP2(Point));
pointcut pointMoveCalls(Point pt):
calls(void pt.setX(int)) || calls(void pt.setY(int));
static after(Object mover, FigureElement movee):
MoveCalls(mover, movee) {
_movees.add(movee);
movee.lastMovedBy = mover;
introduction adds members to target type
public and private are with respect to enclosing aspect declaration
AspectJ Tutorial

71. calls/receptions/executions
differences among :
class MyPoint extends Point {
int getX() { return super.getX(); } }
aspect ShowAccesses {
static before(): calls(void Point.getX()) { <a> }
static before(): receptions(void Point.getX()) { <b> }
static before(): executions(void Point.getX()) { <c> }
code <a> runs once
(new MyPoint()).getX() code <b> runs once
code <c> runs twice
AspectJ Tutorial

72. calls/receptions/executions
differences among :
class MyPoint extends Point {
MyPoint() { ... } }
aspect ShowAccesses {
static before(): calls(Point.new()) { <a> }
static before(): receptions(Point.new()) { <b> }
static before(): executions(Point.new()) { <c> }
remember the implicit super call here!
code <a> runs once
new MyPoint() code <b> runs once
code <c> runs twice
AspectJ Tutorial

73. summary join points method & constructor calls aspects pointcuts
call receptions
executions
field
gets
sets
exception handler
aspects
crosscutting type
of eachobject …class …cflowroot
pointcuts
-primitive-
calls receptions executions handlers gets sets instanceof hasaspect within withincode cflow cflowtop -user-defined- pointcut declaration ‘abstract’
overriding
advice
before
after
around
static non-static of each…
inheritance
introduction
dispatch
AspectJ Tutorial

74. where we have been… … and where we are going problem structure
Part IV:
crosscutting in the design, and how to use AspectJ to capture that
AspectJ mechanisms
Part II:
crosscutting in the code mechanisms AspectJ provides
AspectJ Tutorial

75. AspectJ Tutorial AspectJ IDE support
Part III
AspectJ IDE support
AspectJ Tutorial

76. programming environment
AspectJ Tutorial
programming environment
AJDE support for
emacs, Jbuilder 3.5, Forte 4J
also jdb style debugger (ajdb)
and window-based debugger
navigating AspectJ code
compiling
tracking errors
debugging
ajdoc
AspectJ Tutorial

77. AspectJ Tutorial Using Aspects
Part IV
Using Aspects
AspectJ Tutorial

78. where we have been… … and where we are going problem structure
Part IV:
crosscutting in the design, and how to use AspectJ to capture that
AspectJ mechanisms
Part II:
crosscutting in the code mechanisms AspectJ provides
AspectJ Tutorial

79. goals of this chapter present examples of aspects in design
AspectJ Tutorial
goals of this chapter
present examples of aspects in design
intuitions for identifying aspects
present implementations in AspectJ
how the language support can help
work on implementations in AspectJ
putting AspectJ into practice
raise some style issues
objects vs. aspects
when are aspects appropriate?
AspectJ Tutorial

80. example 1 plug tracing into the system an unpluggable aspect
AspectJ Tutorial
example 1
plug & play tracing
plug tracing into the system
exposes join points and uses very simple advice
an unpluggable aspect
the program’s functionality is unaffected by the aspect
uses both aspect and object
AspectJ Tutorial

81. tracing without AspectJ
AspectJ Tutorial
tracing without AspectJ
class TraceSupport {
static int TRACELEVEL = 0;
static protected PrintStream stream = null;
static protected int callDepth = -1;
static void init(PrintStream _s) {stream=_s;}
static void traceEntry(String str) {
if (TRACELEVEL == 0) return;
callDepth++;
printEntering(str); }
static void traceExit(String str) {
callDepth--;
printExiting(str);
TraceSupport
class Point {
void set(int x, int y) {
TraceSupport.traceEntry(“Point.set”);
_x = x; _y = y;
TraceSupport.traceExit(“Point.set”); }
AspectJ Tutorial

82. a clear crosscutting structure
AspectJ Tutorial
a clear crosscutting structure
all modules of the system use the trace facility in a consistent way:
entering the methods and
exiting the methods
TraceSupport
Say consistent over and over again. What programming in AJ is _understanding_ what is consistent in a crosscut.
hilsdale:
lesson: what it means to be a good aspectJ programmer is being able to identify the _consistency_ or inviariant underlying a crosscut.
what is it about these 12 places? What is the invariant?
gregor: should be what is the invariant that underlies the crosscut
this line is about
interacting with
the trace facility
AspectJ Tutorial

83. tracing as an aspect aspect MyClassTracing { pointcut points():
AspectJ Tutorial
tracing as an aspect
aspect MyClassTracing {
pointcut points():
within(com.bigboxco.boxes.*) &&
executions(* *(..));
static before(): points() {
TraceSupport.traceEntry(
tjp.className + “.” + tjp.methodName); }
static after(): points() {
TraceSupport.traceExit(
TraceSupport
AspectJ Tutorial

84. plug and debug plug in: unplug: or… ajc Point.java Line.java
AspectJ Tutorial
plug and debug
plug in:
unplug:
or…
ajc Point.java Line.java
TraceSupport.java MyClassTracing.java
ajc Point.java Line.java
AspectJ Tutorial

85. plug and debug // log( "New request " + rrequest );
AspectJ Tutorial
plug and debug
//From ContextManager
public void service( Request rrequest, Response rresponse ) {
// log( "New request " + rrequest );
try {
// System.out.print("A");
rrequest.setContextManager( this );
rrequest.setResponse(rresponse);
rresponse.setRequest(rrequest);
// wront request - parsing error
int status=rresponse.getStatus();
if( status < 400 )
status= processRequest( rrequest );
if(status==0)
status=authenticate( rrequest, rresponse );
if(status == 0)
status=authorize( rrequest, rresponse );
if( status == 0 ) {
rrequest.getWrapper().handleRequest(rrequest,
rresponse);
} else {
// something went wrong
handleError( rrequest, rresponse, null, status ); }
} catch (Throwable t) {
handleError( rrequest, rresponse, t, 0 );
// System.out.print("B");
rresponse.finish();
rrequest.recycle();
rresponse.recycle();
} catch( Throwable ex ) {
if(debug>0) log( "Error closing request " + ex);
// log( "Done with request " + rrequest );
// System.out.print("C");
return;
// log( "New request " + rrequest );
// System.out.print(“A”);
// System.out.print("B");
if(debug>0)
log("Error closing request " + ex);
// log("Done with request " + rrequest);
// System.out.print("C");
AspectJ Tutorial

86. plug and debug turn debugging on/off without editing classes
AspectJ Tutorial
plug and debug
turn debugging on/off without editing classes
debugging disabled with no runtime cost
can save debugging code between uses
can be used for profiling, logging
easy to be sure it is off
this connects to point about how to start using aj – compile w/o ajc
AspectJ Tutorial

87. these next two slides seem redundant
AspectJ Tutorial
Gregor Kiczales:
these next two slides seem redundant
aspects in the design
hilsdale:
No, they seem repetetive. They’re here so we can stress design vs. implementation. Perhaps we should emphasize (bold/ital) “responsible for” vs “contains”, etc? Another option would be to “invert” the slides: instead of having 2 slides with three points each (2x3), have 3 slides of 2 points each.
have these benefits
objects are no longer responsible for using the trace facility
trace aspect encapsulates that responsibility, for appropriate objects
if the Trace interface changes, that change is shielded from the objects
only the trace aspect is affected
removing tracing from the design is trivial
just remove the trace aspect
Very important point at this point in talk. When you pick up smalltalk you don’t write algol programs. Your PL changes your thinking. Implications and benefits in the design in addition to in the code.
When you pick up PL1 you do write algol programs.
Very careful with this point: Say “while strong aop is a radical shift, and you can imagine an aop language that would be as jarring as the difference between algol & smalltalk, what we’ve done with aspectJ is this incremental package. We can do this because objects are pretty good”. Don’t say that aspects lead you to a completely different design. It will lead you to think differently, but not in big parts of the design.
AspectJ Tutorial

88. aspects in the code object code contains no calls to trace functions
have these benefits
object code contains no calls to trace functions
trace aspect code encapsulates those calls, for appropriate objects
if the trace interface changes, there is no need to modify the object classes
only the trace aspect class needs to be modified
removing tracing from the application is trivial
compile without the trace aspect class
AspectJ Tutorial

89. tracing: object vs. aspect
AspectJ Tutorial
tracing: object vs. aspect
using an object captures tracing support, but does not capture its consistent usage by other objects
using an aspect captures the consistent usage of the tracing support by the objects
TraceSupport
TraceSupport
make change in speech.
what “consistent” wants to be saying is…
“the structure of the crosscutting”
“the invariant underlying the crosscutting”
AspectJ Tutorial

90. AspectJ Tutorial
tracing
exercises
Make the tracing aspect a library aspect by using an abstract pointcut.
The after advice used runs whether the points returned normally or threw exceptions, but the exception thrown is not traced. Add advice to do so.
AspectJ Tutorial

91. refactor TraceMyClasses into a reusable (library) aspect and an extension equivalent to TraceMyClasses
exercise
aspect TracingXXX {
// what goes here? }
aspect TraceMyClasses extends TracingXXX {
// what goes here? }
AspectJ Tutorial

92. we now have the Trace class, and two aspects, from a design perspective, what does each implement?
exercise
abstract aspect TracingProtocol {
abstract pointcut tracePoints();
static before(): points() {
Trace.traceEntry(tjp.className + “.” + tjp.methodName); }
static after(): points() {
Trace.traceExit(tjp.className + “.” + tjp.methodName);
aspect TraceMyClasses extends TracingProtocol {
pointcut points():
within(com.bigboxco.boxes.*) &&
executions(* *(..)); }
AspectJ Tutorial

93. there are many possible interesting things to do here:
Gregor Kiczales:
there are many possible interesting things to do here:
- teach that you can have a pointcut as shown here
teach that you might fold Trace into Tracing Protocol (call it Tracing)
abstract aspect TracingProtocol {
abstract pointcut points();
pointcut executionPoints(): points() && executions(* *(..));
static before(): executionPoints() {
Trace.traceEntry(tjp.className + “.” + tjp.methodName); }
static after(): executionPoints() {
Trace.traceExit(tjp.className + “.” + tjp.methodName);
aspect TraceMyClasses extends TracingProtocol {
pointcut points():
within(com.bigboxco.boxes.*); }
AspectJ Tutorial

94. example 2
roles/views
AspectJ Tutorial

95. CloneablePoint aspect CloneablePoint { introduction Point {
implements Cloneable;
public Object clone() throws CloneNotSupportedException {
// we choose to bring all fields up to date before cloning
makeRectangular(); // defined in class Point
makePolar(); // defined in class Point
return super.clone(); }
AspectJ Tutorial

96. SerializablePoint aspect SerializablePoint { introduction Point {
implements Serializable;
private static final ObjectStreamField[]
serialPersistentFields = {
new ObjectStreamField("x", Double.TYPE),
new ObjectStreamField("y", Double.TYPE) };
private void writeObject(ObjectOutputStream s)
throws IOException {
makeRectangular();
s.defaultWriteObject(); }
private void readObject(ObjectInputStream s)
throws IOException, ClassNotFoundException {
s.defaultReadObject();
rectangular = true;
polar = false;
AspectJ Tutorial

97. roles/views Write the HashablePoint and ComparablePoint aspects.
AspectJ Tutorial
roles/views
exercises
Write the HashablePoint and ComparablePoint aspects.
Consider a more complex system. Would you want the HashablePoint aspect associated with the Point class, or with other HashableX objects, or both?
Compare using aspects for role/view abstraction with other techniques or patterns.
AspectJ Tutorial

98. example 3 counting bytes interface OutputStream {
public void write(byte b);
public void write(byte[] b); }
/**
* This SIMPLE aspect keeps a global count of all
* all the bytes ever written to an OutputStream. */
aspect ByteCounting {
static int count = 0;
static int getCount() { return count; }
// //
// what goes here? //
AspectJ Tutorial

99. exercise complete the code for ByteCounting /**
* This SIMPLE aspect keeps a global count of all
* all the bytes ever written to an OutputStream. */
aspect ByteCounting {
static int count = 0;
static int getCount() { return count; } }
AspectJ Tutorial

100. counting bytes v1 a first attempt aspect ByteCounting {
static int count = 0;
static int getCount() { return count; }
static after():
receptions(void OutputStream.write(byte)) {
count = count + 1; }
static after(byte[] bytes):
receptions(void OutputStream.write(bytes)) {
count = count + bytes.length;
AspectJ Tutorial

101. counting bytes some stream implementations
AspectJ Tutorial
counting bytes
some stream implementations
class SimpleOutputStream implements OutputStream {
public void write(byte b) { }
public void write(byte[] b) {
for (int i = 0; i < b.length; i++) write(b[i]); }
class OneOutputStream implements OutputStream {
public void write(byte b) {..}
public void write(byte[] b) {..}
AspectJ Tutorial

102. counting bytes another implementation
class OtherOutputStream implements OutputStream {
public void write(byte b) {
byte[] bs = new byte[1];
bs[0] = b;
write(bs); }
public void write(byte[] b) { }
AspectJ Tutorial

103. counting bytes v2 using cflow for more robust counting
aspect ByteCounting {
static int count = 0;
static int getCount() { return count; }
pointcut allWrites(): receptions(void OutputStream.write(byte)) ||
receptions(void OutputStream.write(byte[]));
pointcut withinWrite(): cflow(allWrites());
static after():
!withinWrite() && receptions(void OutputStream .write(byte)) {
count++; }
static after(byte[] bytes):
!withinWrite() && receptions(void OutputStream .write(bytes)) {
count = count + bytes.length;
AspectJ Tutorial

104. counting bytes v3 per-stream counting
aspect ByteCounting of eachobject(allWrites()) {
int count;
static int getCountOf(OutputStream str) {
return ByteCounting.aspectOf(str).count; }
... count++;
... count += bytes.length;
AspectJ Tutorial

105. AspectJ Tutorial
counting bytes
exercises
How would you count bytes written over this interface without aspects?
How do the aspects change if the method void write(Collection c) is added to the OutputStream interface?
Consider a system in which you wrote not only bytes, but byte generators (Objects with a run() method that may output its own bytes). How would you need to change v2?
AspectJ Tutorial

106. example 4 context-passing aspects caller1 caller2 Service
workers need to know the caller:
capabilities
charge backs
to customize result
worker 1
worker 3
worker 2
AspectJ Tutorial

107. context-passing aspects
AspectJ Tutorial
context-passing aspects
caller1
caller2
Service
workers need to know the caller:
capabilities
charge backs
to customize result
sucks that the grey became red
I’ll have to fix this, it MAY be easy, if the rest of the tutorial was using symbolic colors consistently.
worker 1
worker 3
worker 2
AspectJ Tutorial

108. context-passing aspects
pointcut invocations(Caller c):
instanceof(c) && calls(void Service.doService(String));
AspectJ Tutorial

109. context-passing aspects
pointcut invocations(Caller c):
instanceof(c) && calls(void Service.doService(String));
pointcut workPoints(Worker w):
receptions(void w.doTask(Task));
AspectJ Tutorial

110. context-passing aspects
pointcut invocations(Caller c):
instanceof(c) && calls(void Service.doService(String));
pointcut workPoints(Worker w):
receptions(void w.doTask(Task));
pointcut perCallerWork(Caller c, Worker w):
cflow(invocations(c)) && workPoints(w);
AspectJ Tutorial

111. context-passing aspects
abstract aspect CapabilityChecking {
pointcut invocations(Caller c):
instanceof(c) && calls(void Service.doService(String));
pointcut workPoints(Worker w):
receptions(void w.doTask(Task));
pointcut perCallerWork(Caller c, Worker w):
cflow(invocations(c)) && workPoints(w);
before (Caller c, Worker w): perCallerWork(c, w) {
w.checkCapabilities(c); }
AspectJ Tutorial

112. context-passing aspects
AspectJ Tutorial
context-passing aspects
exercises
The before advice on the perCallerWork pointcut calls the worker’s checkCapabilities method to check the capabilities of the caller. What would be an appropriate way to write that method?
AspectJ Tutorial

113. example 5 properties of interfaces interface Forest {
AspectJ Tutorial
example 5
properties of interfaces
interface Forest {
int howManyTrees();
int howManyBirds();
... }
pointcut forestReceptions(): receptions(* Forest.*(..));
before(): forestReceptions(): {
AspectJ Tutorial

114. aspects on interfaces a first attempt aspect Forestry {
AspectJ Tutorial
aspects on interfaces
a first attempt
aspect Forestry {
pointcut ForestMethodReceptions():
receptions(* Forest.*(..));
static before(): ForestMethodReceptions() {
System.out.println(thisJoinPoint.methodName +
" is a Forest-method."); }
AspectJ Tutorial

115. aspects on interfaces an implementation
AspectJ Tutorial
aspects on interfaces
an implementation
class ForestImpl implements Forest {
public static void main(String[] args) {
Forest f1 = new ForestImpl();
f1.toString();
f1.howManyTrees(); }
public int howManyTrees() { return 100; }
public int howManyBirds() { return 200; }
interface Forest includes methods from Object, such as toString()
AspectJ Tutorial

116. aspects on interfaces adding constraints aspect Forestry {
AspectJ Tutorial
aspects on interfaces
adding constraints
aspect Forestry {
pointcut ForestMethodReceptions():
receptions(* Forest.*(..)) &&
!receptions(* Object.*(..));
static before(): ForestMethodReceptions() {
System.out.println(thisJoinPoint.methodName +
" is a Forest-method."); }
AspectJ Tutorial

117. AspectJ Tutorial
aspects on interfaces
exercises
In this example you needed to constrain a pointcut because of undesired inheritance. Think of an example where you would want to capture methods in a super-interface.
Constraining a pointcut in this way can be seen as an aspect idiom. What other idioms have you seen in this tutorial?
AspectJ Tutorial

118. example 6 client reactions to failures: - abort - try another server
RMI exception aspects
client reactions to failures:
- abort
- try another server
AspectJ Tutorial

119. a TimeServer design
AspectJ Tutorial

120. the TimeService public interface TimeService extends Remote { /**
* What's the time? */
public Date getTime() throws RemoteException;
* Get the name of the server
public String getName() throws RemoteException;
* Exported base name for the service
public static final String nameBase = "TimeService"; }
AspectJ Tutorial

121. no exception catching here, but notice
AspectJ Tutorial
the TimeServer
public class TimeServer extends UnicastRemoteObject
implements TimeService {
/**
* The remotely accessible methods */
public Date getTime() throws RemoteException {return new Date();}
public String getName() throws RemoteException {return toString();}
* Make a new server object and register it
public static void main(String[] args) {
TimeServer ts = new TimeServer();
Naming.bind(TimeService.nameBase, ts); }
* Exception pointcuts. Code won’t compile without advising them
pointcut create() returns TimeServer:
within(TimeServer) && calls(TimeServer.new());
pointcut bind(String name) returns void:
within(TimeServer) && calls(void Naming.bind(name,..));
Gregor Kiczales:
people liked this – it teaches using AspectJ to expose an internal interface
no exception catching here, but notice
Some objects you can take out and the application still runs, many of them you can’t. Aspects are the same way.
Objects are a modularity mechanism, and when you use them … Aspects are the same. The only difference is that Aspects are about crosscutting. That’s the only difference.
AspectJ Tutorial

122. AbortMyServer aspect AbortMyServer {
static around() returns TimeServer: TimeServer.create() {
TimeServer result = null;
try {
result = thisJoinPoint.runNext();
} catch (RemoteException e){
System.out.println("TimeServer err: " + e.getMessage());
System.exit(2); }
return result;
static around(String name) returns void: TimeServer.bind(name) {
thisJoinPoint.runNext(name);
System.out.println("TimeServer: bound name.");
} catch (Exception e) {
System.err.println("TimeServer: error " + e);
System.exit(1);
AspectJ Tutorial

123. RetryMyServer aspect RetryMyServer {
static around() returns TimeServer: TimeServer.create() {
TimeServer result = null;
try { result = thisJoinPoint.runNext(); }
catch (RemoteException e){
System.out.println("TimeServer error."); e.printStackTrace(); }
return result;
static around(String name) returns void: TimeServer.bind(name) {
for (int tries = 0; tries < 3; tries++) {
try {
thisJoinPoint.runNext(name + tries);
System.out.println("TimeServer: Name bound in registry.");
return;
} catch (AlreadyBoundException e) {
System.err.println("TimeServer: name already bound");
System.err.println("TimeServer: Giving up."); System.exit(1);
AspectJ Tutorial

124. again, no exception catching here
the Client
again, no exception catching here
public class Client {
TimeService server = null;
/**
* Get a server and ask it the time occasionally */
void run() {
server = (TimeService)Naming.lookup(TimeService.nameBase);
System.out.println("\nRemote Server=" + server.getName() + "\n\n");
while (true) {
System.out.println("Time: " + server.getTime());
pause(); }
* Exception pointcuts. Code won’t compile without advising them
pointcut setup(Client c) returns Remote:
instanceof(c) & calls(Remote Naming.lookup(..));
pointcut serve(Client c, TimeService ts) returns Object:
instanceof(c) & calls(* ts.*(..));
… other methods …
AspectJ Tutorial

125. AbortMyClient aspect AbortMyClient {
static around(Client c) returns Remote: Client.setup(c) {
Remote result = null;
try {
result = thisJoinPoint.runNext(c);
} catch (Exception e) {
System.out.println("Client: No server. Aborting.");
System.exit(0); }
return result;
static around(Client c, TimeService ts) returns Object:
Client.serve(c, ts) {
Object result = null;
result = thisJoinPoint.runNext(c, ts);
} catch (RemoteException e) {
System.out.println("Client: Remote Exception. Aborting.");
AspectJ Tutorial

126. RetryMyClient aspect RetryMyClient {
static around(Client c) returns Remote: Client.setup(c) {
Remote result = null;
try { result = thisJoinPoint.runNext(c);}
catch (NotBoundException e) {
System.out.println("Client: Trying alternative name...");
result = findNewServer(TimeService.nameBase, c.server, 3);
if (result == null) System.exit(1); /* No server found */
} catch (Exception e2) { System.exit(2); }
return result; }
static around(Client c, TimeService ts) returns Object:
Client.serve(c,ts) {
try { return thisJoinPoint.runNext(c,ts); }
catch (RemoteException e) { /* Ignore and try other servers */ }
c.server = findNewServer(TimeService.nameBase, c.server, 3);
if (c.server == null) System.exit(1); /* No server found */
try { return thisJoinPoint.runNext(c, c.server); }
catch (RemoteException e2) { System.exit(2); }
return null;
static TimeService findNewServer(String baseName,
Object currentServer, int nservers) { … }
AspectJ Tutorial

127. building the client abort mode: retry mode:
AspectJ Tutorial
building the client
abort mode:
retry mode:
switch to different failure handling modes without editing
no need for subclassing or delegation
reusable failure handlers
ajc Client.java TimeServer_Stub.java AbortMyClient.java
ajc Client.java TimeServer_Stub.java RetryMyClient.java
somewhere should say
“the benefits are the usual benefits of better modularity” the difference is that aspectj makes it possible to modularize crosscutting concerns -- aspects
AspectJ Tutorial

128. RMI exception handling
AspectJ Tutorial
RMI exception handling
exercises
Write another exception handler that, on exceptions, gives up the remote mode and instantiates a local TimeServer.
How would this client look like if the exception handling were not designed with aspects? Can you come up with a flexible OO design for easily switching between exception handlers?
Compare the design of exception handlers with aspects vs. with your OO design
AspectJ Tutorial

129. example 7
layers of functionality
given a basic telecom operation, with customers, calls, connections
model/design/implement utilities such as
timing
consistency checks
...
AspectJ Tutorial

130. telecom basic design Customer Call Connection Local LongDistance
Call startCall(Customer)
void pickupCall(Call)
void mergeCalls(Call, Call)
void hangupCall(Call)
Call
Customer getCaller()
void pickup()
void merge(Call)
void hangup()
Connection
Customer getReceiver()
void complete()
void drop()
0..N 1
caller
receiver
Local
LongDistance
These classes
define the
protocols for
setting up calls
(includes
conference
calling) and
establishing
connections
AspectJ Tutorial

131. timing entities Call Customer Connection Local LongDistance
store total
connection
time
Call
Customer 1 1 1
0..N
caller
Connection
time each connection
receiver
Local
LongDistance
AspectJ Tutorial

132. timing some actions Call Customer Connection Local LongDistance
connection dropped:
add time
Customer 1 1 1
0..N
connection made:
start timing
connection dropped:
stop timing
caller
Connection
receiver
Local
LongDistance
AspectJ Tutorial

133. this piece of setup is inconsistent with the code
timing
additional design elements
0..N
long getTime()
void addToTime(long t)
Call
Customer 1
Gregor Kiczales:
this piece of setup is inconsistent with the code 1 1
0..N
Timer
void stop()
void start()
long getTime()
invoke when
connection drops
caller
Connection
set and invoke
upon new
connection
receiver
Local
LongDistance
AspectJ Tutorial

134. timing Write an aspect representing the timing protocol. exercise
AspectJ Tutorial

135. timing connections and calls are involved
what is the nature of the crosscutting?
connections and calls are involved
well defined protocols among them
pieces of the timing protocol must be triggered by the execution of certain basic operations. e.g.
when connection is completed, set and start a timer
when connection drops, stop the timer and add time to customers’ connection time
AspectJ Tutorial

136. timing an aspect implementation aspect Timing {
static aspect ConnectionTiming of eachobject(instanceof(Connection)) {
private Timer timer = new Timer(); }
static aspect CustomerTiming of eachobject(instanceof(Customer)) {
private long totalConnectTime = 0;
public long getTotalConnectTime() {
return totalConnectTime;
pointcut startTiming(Connection c): receptions(void c.complete());
pointcut endTiming(Connection c): receptions(void c.drop());
static after(Connection c): startTiming(c) {
ConnectionTiming.aspectOf(c).timer.start();
static after(Connection c): endTiming(c) {
Timer timer = ConnectionTiming.aspectOf(c).timer;
timer.stop();
long currTime = timer.getTime();
CustomerTiming.aspectOf(c.getCaller()).totalConnectTime += currTime;
CustomerTiming.aspectOf(c.getReceiver()).totalConnectTime += currTime;
AspectJ Tutorial

137. addToTime(timer.getTime())
timing as an object
Customer
...
long getTime()
void addToTime(long)
0..N 1
addToTime(timer.getTime())
0..N
Timer
void stop()
void start()
long getTime()
Connection
...
drop()
new(..)
timing as an object captures timing support, but does not capture the protocols involved in implementing the timing feature
AspectJ Tutorial

138. addToTime(timer.getTime())
timing as an aspect
Timing
Customer
...
0..N
long getTime()
void addToTime(long t) 1
addToTime(timer.getTime())
0..N
Timer
void stop()
void start()
long getTime()
Connection
...
drop()
new(..)
timing as an aspect captures the protocols involved in implementing the timing feature
AspectJ Tutorial

139. timing as an aspect
has these benefits
basic objects are not responsible for using the timing facility
timing aspect encapsulates that responsibility, for appropriate objects
if requirements for timing facility change, that change is shielded from the objects
only the timing aspect is affected
removing timing from the design is trivial
just remove the timing aspect
AspectJ Tutorial

140. timing with AspectJ object code contains no calls to timing functions
has these benefits
object code contains no calls to timing functions
timing aspect code encapsulates those calls, for appropriate objects
if requirements for timing facility change, there is no need to modify the object classes
only the timing aspect class and auxiliary classes needs to be modified
removing timing from the application is trivial
compile without the timing aspect class
AspectJ Tutorial

141. AspectJ Tutorial
timing
exercises
How would you change your program if the interface to Timer objects changed to
What changes would be necessary without the aspect abstraction?
Timer
void start()
long stopAndGetTime()
AspectJ Tutorial

142. telecom, continued
layers of functionality: consistency
ensure that all calls and connections are being shut down in the simulation
AspectJ Tutorial

143. consistency checking Call List of Calls Connection List of Connections
AspectJ Tutorial
consistency checking
store each new call
remove calls that
hangup
merge
Call
Customer getCaller()
void pickup()
void merge(Call)
void hangup()
Customer
Call startCall(Customer)
void pickupCall(Call)
void mergeCalls(Call, Call)
void hangupCall(Call) 1
0..N
List of
Calls 1 1
0..N
0..N
Connection
Customer getCaller()
Customer getReceiver()
void complete()
void drop()
store each new connection
remove connections that drop
caller
I put a small background-colored box in the lower left corner of this slide to fix its appearance.
receiver
List of
Connections
0..N
Local
LongDistance
check at the end
of simulation
AspectJ Tutorial

144. consistency checking aspect ConsistencyChecker {
static Vector calls = new Vector(), connections = new Vector();
/* The lifecycle of calls */
static after(Call c): receptions(c.new(..)) {
calls.addElement(c); }
static after(Call c): receptions(* c.hangup(..)) {
calls.removeElement(c);
static after(Call other): receptions(void Call.merge(other)) {
calls.removeElement(other);
/* The lifecycle of connections */
static after(Connection c): receptions(c.new(..)) {
connections.addElement(c);
static after(Connection c): receptions(* c.drop(..)) {
connections.removeElement(c);
static after(): within(TelecomDemo) && executions(void main(..)) {
if (calls.size() != 0) println("ERROR on calls clean up.");
if (connections.size()!=0) println("ERROR on connections clean up.");
AspectJ Tutorial

145. summary so far presented examples of aspects in design
AspectJ Tutorial
summary so far
presented examples of aspects in design
intuitions for identifying aspects
presented implementations in AspectJ
how the language support can help
raised some style issues
objects vs. aspects
AspectJ Tutorial

146. when are aspects appropriate?
AspectJ Tutorial
when are aspects appropriate?
is there a concern that:
crosscuts the structure of several objects or operations
is beneficial to separate out
hilsdale:
in wrap=up part, do wrapup based on stuff
AspectJ Tutorial

147. … crosscutting
a design concern that involves several objects or operations
implemented without AOP would lead to distant places in the code that
do the same thing
e.g. traceEntry(“Point.set”)
try grep to find these [Griswold]
do a coordinated single thing
e.g. timing, observer pattern
harder to find these
AspectJ Tutorial

148. … beneficial to separate out
AspectJ Tutorial
… beneficial to separate out
does it improve the code in real ways?
separation of concerns
e.g . think about service without timing
clarifies interactions, reduces tangling
e.g. all the traceEntry are really the same
easier to modify / extend
e.g. change the implementation of tracing
e.g. abstract aspect re-use
plug and play
tracing aspects unplugged but not deleted
standard q: what If people use this too much. A: they will, and then they’ll learn. It’s like any new feature. Good taste is learned from experience. What we’re trying to teach is how to use aspects when you need it. If you don’t, please don’t.
hilsdale:
NOT HERE:
simple d: of crosscutting:
You have it if with respect to one concern, one element is doing one thing, but with respect to another concern, one element is doing n things. Or vice versa.
Don’t use this.
AspectJ Tutorial

149. learned through experience, influenced by taste and style
good designs
summary
capture “the story” well
may lead to good implementations, measured by
code size
tangling
coupling
etc.
learned through experience, influenced by taste and style
AspectJ Tutorial

150. expected benefits of using AOP
AspectJ Tutorial
expected benefits of using AOP
good modularity, even in the presence of crosscutting concerns
less tangled code, more natural code, smaller code
easier maintenance and evolution
easier to reason about, debug, change
more reusable
more possibilities for plug and play
abstract aspects
AspectJ Tutorial

151. References, Related Work
Part V
References, Related Work
AspectJ Tutorial

152. AOP and AspectJ on the web
aspectj.org
AspectJ Tutorial

153. Workshops ECOOP’97 ICSE’98 ECOOP’98 ECOOP’99 OOPSLA’99 ECOOP’00
AspectJ Tutorial
Workshops
ECOOP’97
ICSE’98
ECOOP’98
ECOOP’99
OOPSLA’99
ECOOP’00
OOPSLA’00
AspectJ Tutorial

154. growing interest aspect-oriented programming
in separation of crosscutting concerns
aspect-oriented programming
composition U Twente
[Aksit]
adaptive Northeastern U
[Lieberherr]
multi-dimensional separation of IBM
[Ossher, Tarr]
assessment of SE UBC
[Murphy]
information UCSD
[Griswold] …
AspectJ Tutorial

155. AOP future – idea, language, tools
AspectJ Tutorial
AOP future – idea, language, tools
objects are
code and state
“little computers”
message as goal
hierarchical structure
languages support
encapsulation
polymorphism
inheritance
tools
browser, editor, debuggers
preserve object abstraction
aspects are
crosscutting structure
languages support
crosscutting
tools
preserve aspect abstraction
I’m going to show you a very particular language. As you look at it, think Smalltalk, CLOS, Java. I.e. think about both it and the space around it.
AspectJ Tutorial

156. AOP future language design tools software engineering metrics theory
AspectJ Tutorial
AOP future
language design
more dynamic crosscuts, type system …
tools
more IDE support, aspect discovery, re-factoring, re-cutting…
software engineering
finding aspects, modularity principles, …
metrics
measurable benefits, areas for improvement
theory
type system for crosscutting, fast compilation, advanced crosscut constructs
AspectJ Tutorial

157. AspectJ & the Java platform
AspectJ is a small extension to the Java programming language
all valid programs written in the Java programming language are also valid programs in the AspectJ programming language
AspectJ has its own compiler, ajc
ajc runs on Java 2 platform
ajc is available under Open Source license
ajc produces Java platform compatible .class files
AspectJ Tutorial

158. AspectJ status release status aspectj.org tutorials & training tools
3 major, ~18 minor releases over last year (0.7beta8 is current)
tools
IDE extensions: Emacs, JBuilder 3.5, Forte4J
ajdoc to parallel javadoc
debugger: command line, GUI, & IDE
license
compiler, runtime and tools are free for any use
compiler and tools are Open Source
aspectj.org
May 1999: 90 downloads/mo, 20 members on users list
May 2000: 500 downloads/mo, 500 members on users list
tutorials & training
3 tutorials in 1999, 8 in 1999, 12 in 2000
Gregor Kiczales:
catch dates up
AspectJ Tutorial

159. AspectJ future continue building language, compiler & tools 0.8 0.9
one more IDE
more powerful join point designators
0.9
inheritance, a few other language cleanups
1.0
minor language tuning
incremental compilation, compilation to bytecodes
at least two more IDEs
1.1
faster incremental compiler (up to 5k classes)
source of target classes not required
2.0
new dynamic crosscut constructs
commercialization decision after 1.0
Gregor Kiczales:
catch this up to what we said at OOPSLA, I think, for example, that first version incremental is coming sooner than this says
AspectJ Tutorial

160. credits AspectJ.org is a Xerox PARC project:
Bill Griswold, Erik Hilsdale, Jim Hugunin, Mik Kersten, Gregor Kiczales, Jeffrey Palm
slides, compiler, tools & documentation are available at aspectj.org
partially funded by DARPA under contract F C0246
AspectJ Tutorial