One of the issues reported with BTrace is that the trace authors have to write "verbose" code [some people say Java is "verbose"!]. In BTrace, we have to repeat the same set of imports, annotations in every btrace file and all methods have to be "public static void" and so on. Instead of inventing a new scripting language, I've added a simple C preprocessor like step in BTrace compilation. This preprocessor is based on the one in the GlueGen project. Thanks to Ken Russell for this code and for reviewing my changes specific to BTrace project. The preprocessor solution does not rule out a scripting solution in future
If you have nice ideas or would like contribute in this area, you are always welcome! But, I think preprocessor solution is simple and will be useful to some.
Simple Example:
btracedefs.h
ThreadBean.java
To run this sample, the following command can be used:
btrace -I . <pid-of-the-traced-process> ThreadBean.java
Without the -I option in command line, BTrace skips the preprocessor step.
Squeak is a open source implementation of Smalltalk. What is JSqueak? JSqueak is a Squeak interpreter written in Java. You can download JSqueak source code and play with it. I did the following:
This is how it looks...
You can dyanamically attach BTrace to a Java process to inject trace code into it. BTrace client classescollect the trace output via a socket -- these client classes are used by BTrace command line client as well as VisualVM plugin for BTrace. How about attaching a JMX client to collect BTrace's trace data? Yes, it is possible to access a BTrace class's static fields as attributes of a MBean with this RFE.
There are two MBean samples in the BTrace repository. I attached both BTrace samples to a "Java2D demo" process. And then I attached VisualVM to view the Mbean registered by these BTrace samples:
I work from home in Chennai, India. There is maintenance power shutdown in my part of the city today [from 9.00 AM to 5.00 PM). I'm writing this blog from a Sun office in Apeejay Business Centre, Chennai. It is nice to be in an office after quite some time - at least as a change! But, I think I'd rather prefer to avoid travel, preparation to go office etc. every day
If you have used DTrace, chances are that you have used aggregations. For performance issues, aggregated data is often more useful than individual data points. With BTrace, aggregating data is bit painful (you have to manage using Maps explicitly). It would be nice to have DTrace-style aggregation functions such as sum, max, min and so on. Glencross, Christian M (cited in my previous entry) has contributed code changes, doc and a sample for easy-to-use aggregation facility for BTrace. Please refer to the sample code (JdbcQueries.java) that demonstrates aggregations.
Now something unrelated to aggregations, but related to BTrace : I came to know about another use-case of BTrace. See also http://blog.igorminar.com/2008/06/btrace-dtrace-for-java.html
In the last few weeks, I came to know about two cases of real world use of BTrace.
import static com.sun.btrace.BTraceUtils.*;
import java.sql.Statement;
import java.util.Map;
import java.util.concurrent.atomic.AtomicLong;
import com.sun.btrace.*;
import com.sun.btrace.annotations.*;
/**
* BTrace script to print timings for all executed JDBC statements on an event.
* <p>
*
* @author Chris Glencross
*/
@BTrace
public class JdbcQueries {
private static Map preparedStatementDescriptions = newWeakMap();
private static Map statementDurations = newHashMap();
// VERBOSE: @TLS makes the field "thread local" -- sort of like using java.lang.ThreadLocal
@TLS
private static String preparingStatement;
@TLS
private static long timeStampNanos;
@TLS
private static String executingStatement;
/**
* If "--stack" is passed on command line, print the Java stack trace of the JDBC statement.
*
* VERBOSE: Command line arguments to BTrace are accessed as $(N) where N is the command line arg position.
*
* Otherwise we print the SQL.
*/
private static boolean useStackTrace = $(2) != null && strcmp("--stack", $(2)) == 0;
// The first couple of probes capture whenever prepared statement and callable statements are
// instantiated, in order to let us track what SQL they contain.
/**
* Capture SQL used to create prepared statements.
*
* VERBOSE: +foo in clazz means foo and it's subtypes. Note the use of regular expression
* for method names. With that BTrace matches all methods starting with "prepare". The
* type "AnyType" matches any Java type.
*
* @param args - the list of method parameters. args[1] is the SQL.
*/
@OnMethod(clazz = "+java.sql.Connection", method = "/prepare.*/")
public static void onPrepare(AnyType[] args) {
preparingStatement = useStackTrace ? jstackStr() : str(args[1]);
}
/**
* Cache SQL associated with a prepared statement.
*
* VERBOSE: By default, @OnMethod matches method entry points. Modifying with @Location
* annotation to match the method return points.
*
* @param arg - the return value from the prepareXxx() method.
*/
@OnMethod(clazz = "+java.sql.Connection", method = "/prepare.*/", location = @Location(Kind.RETURN))
public static void onPrepareReturn(AnyType arg) {
if (preparingStatement != null) {
print("P"); // Debug Prepared
Statement preparedStatement = (Statement) arg;
put(preparedStatementDescriptions, preparedStatement, preparingStatement);
preparingStatement = null;
}
}
// The next couple of probes intercept the execution of a statement. If it execute with no-args,
// then it must be a prepared statement or callable statement. Get the SQL from the probes up above.
// Otherwise the SQL is in the first argument.
@OnMethod(clazz = "+java.sql.Statement", method = "/execute.*/")
public static void onExecute(AnyType[] args) {
timeStampNanos = timeNanos();
if (args.length == 1) {
// No SQL argument; lookup the SQL from the prepared statement
Statement currentStatement = (Statement) args[0]; // this
executingStatement = get(preparedStatementDescriptions, currentStatement);
} else {
// Direct SQL in the first argument
executingStatement = useStackTrace ? jstackStr() : str(args[1]);
}
}
@OnMethod(clazz = "+java.sql.Statement", method = "/execute.*/", location = @Location(Kind.RETURN))
public static void onExecuteReturn() {
if (executingStatement == null) {
return;
}
print("X"); // Debug Executed
long durationMicros = (timeNanos() - timeStampNanos) / 1000;
AtomicLong ai = get(statementDurations, executingStatement);
if (ai == null) {
ai = newAtomicLong(durationMicros);
put(statementDurations, executingStatement, ai);
} else {
addAndGet(ai, durationMicros);
}
executingStatement = null;
}
// VERBOSE: @OnEvent probe fires whenever BTrace client sends "event" command.
// The command line BTrace client sends BTrace events when user pressed Ctrl-C
// (more precisely, on receiving SIGINT signal)
@OnEvent
public static void onEvent() {
println("---------------------------------------------");
printNumberMap("JDBC statement executions / microseconds:", statementDurations);
println("---------------------------------------------");
}
}
And he has expressed few wish lists for BTrace based on his experience with DTrace. We plan to investigate those items in near future.
I received emails asking for BTrace BOF (JavaOne-2008) slides. Better late than never... I've uploaded PDF of the slides. The BOF was mostly around demos -- slides do not contain much. But, slides have few pointers that may be useful.
Here are the few highlights from the talks that I attended today:
TS-5428 Java Technology Meets the Real World: Intelligence Everywhere.
This talk is about pervasive computing (a.k.a ubiquitous computing) with products from Sentilla. There was an interesting demo about humidity sensor detecting changes and sending a message to a host. The "motes" run CLDC 1.1 VM (+ proprietary profile for motes). These motes have ports for sensors and actuators and some built-in sensor. There were many interesting suggestions for embedded programming for such small devices (don't allocate in inner loops and there by leading to to GC kick-in, avoid too many static fields, avoid threads whenever possible and so on).
TS-7575 Using Java Technology-Based Class Loaders to design and implementing a Java platform, Micro Edition
The basic idea is to run JavaME applications (developed for different configurations/profiles/subsets of optional packages) on top of JavaSE. The extended JavaSE classes and packages not available in specific profile or optional package set [implemented by a specific phone] should not be made available to JavaME apps targeted. i.e., only the classes available to a specific phone model should be available. If the JavaME app tries to access any other class, it should receive ClassNotFoundException. The speakers explained how to achieve such "containers" by class loader based isolation. The problem is that they seem to solve only the class access. What about extended methods and fields? For example, platform core classes on JavaSE have superset of methods [more methods on the same class available on JavaME - eg. java.util.Hashtable has more methods on JavaSE). The application classes have to bytecode analyzed and instrumented to take care of field/method accces. It seems that their current product that does not address this yet.
PAN-5542 Developing Semantic Web Applications on the Java Platform.
The discussion started with some nice demos. There was a demo with AllegroGraph RDF store, Twine, a demo with using GRDDL and getting RDF triples by a proxy server. i.e., a proxy serves does the GRDDL transformations to get RDF triples from sites [which could be stored/analyzed with RDF stores subsequently] and a demo with FOAF files. Interesting take aways from the discussion include:
Today Bill, Chihiro, Jaya and I talked on Blu-ray. The talk was centered around the open source project @ http://hdcookbook.dev.java.net - a library and a set of tools to build Blu-ray discs. If you haven't checked out code/docs, you may want to checkout and play with the code. All you need is a laptop with blu-ray drive and a BD-RE disc. Optionally, for added fun you may want to have a hardware bluray player such as PS3 -- so that you can see the output on your TV rather than on a laptop. Other than the session, we also had a very informal BOF on blu-ray, OCAP etc. during the evening. It is good to meet experts in respective technologies in one place!
Other than the the blu-ray stuff, I did attend other talks/BOF. Just after Blu-ray session, I attended "TS-6000 Improving Application Performance with Monitoring and Profiling Tools" talk. This talk was about OS specific tools, JDK tools and third-party tools for profiling and monitoring. Gregg Sporar and Jaroslav Bachorik (NetBeans Profiler team) presented very well. There were many interesting questions/discussions as well. If you haven't done so already, you may want to download VisualVM. If you want bit more fun doing monitoring/profiling, you may want to check out the sources from http://visualvm.dev.java.net and build it yourself. You can build BTrace VisualVM plugin using the command:
c:\visualvm\plugins>ant build
assuming you have checked out VisualVM sources under "c:\visualvm". If you have already checked out BTrace sources under some other directory, say "c:\btrace", you can use
c:\visualvm\plugins>ant -Dbtrace.home=c:\btrace build
To run VisualVM with all the plugins that you built, you can use the following command:
c:\visualvm\plugins>ant -Dbtrace.home=c:\btrace run
Please let us know what features you'd like to see with BTrace and/or BTrace VisualVM plugin.
I attended and liked the "Class Loader Rearchitected (BOF-6180)" BOF. If you have ever written class loaders, chances are that you have faced mysterious deadlocks or ClassCastException that said "ClassCastException: Foo cannot be cast to Foo" or having to decide between overriding loadClass and findclass, you probably should have attended this talk and gave your opinions/suggestions/ideas
If I understood properly, I think there was a suggestion to add class loader info. to the ClassCastException (something like class-loader-class-name@identity-HashCode style string?) so that one can quickly see it is a class loader issue. Also, there were many questions on loading classes from jar files. Looks like there will be changes to class loader API and class loading in VM for JDK 7.
In today's sessions that I attended I liked the following:
JRuby: Why, What, How... Do It Now
This talk is a good introduction to (J)Ruby the language and important applications of (J)Ruby. And many pointers to related (J)Ruby sessions. Nice summary!
JavaScript programming language: The Language Everybody Loves to Hate
great talk by Roberto Chinnici. Nice summary of functional and prototype-based object orientation aspects of JavaScript. You can easily impress your friends will some neat snippets of JavaScript
You may want to continue the fun by reading Doug Crockford's pages, if you have not do already!
At 7.30 PM, we (I and Kannan) talked about BTrace. There were many interesting questions/discussions -- both during and after the BOF! Today (Wed May 7) will be a Blu-ray day -- it starts with TS-5449 Java Technology for Blu-ray and TV: Creating your own Blu-ray Java Discs session. It is about the open source project @ http://hdcookbook.dev.java.net. Meet you all there!
In JavaOne 2008, there are many intesting sessions on "other" JVM languages covering both dynamically typed languages (JavaScript, Groovy, JRuby) and statically typed languages (JavaFX, Scala). As usual, there are many sessions covering application aspects -- like using scripting on Glassfish, Grials (Groovy), Rails (JRuby) and so on. But, my interest is mostly on the programming language aspects and JVM implementation issues. Here is a table of sessions covering those:
|
Session ID |
Session Title |
Session Type |
Speakers and Company |
Speakers and Company |
Venue - Room |
|
|
|
||||||
|
TS-5152 |
Overview of the JavaFX™ Script Programming Language |
Technical Session |
Christopher Oliver, Sun Microsystems, Inc. |
Tuesday |
Moscone Center - |
|
|
TS-5416 |
JRuby: Why, What, How...Do It Now |
Technical Session |
Thomas Enebo, Sun Microsystems, Inc. ; Charles Nutter, Sun Microsystems, Inc. |
Tuesday |
Moscone Center - |
|
|
TS-4794 |
A JavaFX™ Script Programming Language Tutorial |
Technical Session |
James Weaver, LAT |
Tuesday |
Moscone Center - |
|
|
TS-4986 |
JavaScript™ Programming Language: The Language Everybody Loves to Hate |
Technical Session |
Roberto Chinnici, Sun Microsystems, Inc. |
Tuesday |
Moscone Center - |
|
|
PAN-5435 |
The Script Bowl: A Rapid-Fire Comparison of Scripting Languages |
Panel Session |
Guillaume Laforge, G2One, Inc.; Charles Nutter, Sun Microsystems, Inc. ; Jorge Ortiz, Stanford; Raghavan Srinivas, Sun Microsystems, Inc.; Frank Wierzbicki, Sun Microsystems |
Wednesday |
Moscone Center - |
|
|
TS-5572 |
Groovy, the Red Pill: Metaprogramming--How to Blow the Mind of Developers on the Java™ Platform |
Technical Session |
Scott Davis, Davisworld Consulting, Inc. |
Wednesday |
Moscone Center - |
|
|
TS-5165 |
Programming with Functional Objects in Scala |
Technical Session |
Martin Odersky, EPFL |
Thursday |
Moscone Center - |
|
|
TS-5693 |
Writing Your Own JSR-Compliant, Domain-Specific Scripting Language |
Technical Session |
John Colosi, VeriSign, Inc.; David Smith, VeriSign Inc. |
Thursday |
Moscone Center - |
|
|
TS-6050 |
Comparing JRuby and Groovy |
Technical Session |
Neal Ford, ThoughtWorks Inc. |
Friday |
Moscone Center - |
|
|
TS-6039 |
Jython - Implementing Dynamic Language Features for the Java™ Platform Ecosystem |
Technical Session |
Jim Baker, Zyasoft; Tobias Ivarsson, Neo Technology |
Friday |
Moscone Center - |
|
We have a BOF on BTrace in this year's JavaOne. But, you will not find the name "BTrace" in session title -- that is because talk was submitted before BTrace was open sourced with that name
The details of the BOF is as below. Please visit and let us discuss on dynamic tracing for Java.
| BOF-5552 | Java™ Platform Observability by Bytecode Instrumentation | Kannan Balasubramainan, A. Sundararajan | Tuesday May 06 19:30 - 20:20 | Moscone Center - Esplanade 300 |
| TS-5716 | D-I-Y (Diagnose-It-Yourself): Adaptive Monitoring for Sun Java™ Real-Time System | Technical Session | Carlos Lucasius, Frederic Parain | Tuesday May 06 18:00 - 19:00 | Moscone Center - Hall E 133|
| TS-6000 | Improving Application Performance with Monitoring and Profiling Tools | Technical Session | Jaroslav Bachorik, Gregg Sporar | Wednesday May 07 10:50 - 11:50 | Moscone Center - Gateway 104 |
| LAB-9400 | Exposing the Depth of Your JDK™ Release 7.0 Applications with Dynamic Tracing (DTrace) | Hands-On Lab | Angelo Rajadurai, Raghavan Srinivas, | Wednesday May 07 18:30 - 20:30 | Moscone Center - Hall E 130/131 (LAB) |
| TS-6145 | Using DTrace with Java™ Technology-Based Applications: Bridging the Observability Gap | Technical Session | Jonathan Haslam, Simon Ritter | Thursday May 08 13:30 - 14:30 | Moscone Center - North Mtg-121/122/124/125 |
| BOF-4994 | End-to-End Tracing of Ajax/Java™ Technology-Based Applications, Using Dynamic Tracing (dTrace) | Birds-of-a-Feather Session (BOF) | Amit Hurvitz | Thursday May 08 18:30 - 19:20 | Moscone Center - Gateway 104 |
| BOF-5223 | VisualVM: Integrated and Extensible Troubleshooting Tool for the Java™ Platform | Birds-of-a-Feather Session (BOF) | Luis-Miguel Alventosa, Tomas Hurka | Thursday May 08 19:30 - 20:20 | Moscone Center - Gateway 104 |
| TS-6145 | Using DTrace with Java™ Technology-Based Applications: Bridging the Observability Gap | Technical Session | Jonathan Haslam, Simon Ritter | Friday May 09 14:50 - 15:50 | Moscone Center - North Mtg-121/122/124/125 |
| TS-6000 | Improving Application Performance with Monitoring and Profiling Tools | Technical Session | Jaroslav Bachorik, Gregg Sporar | Friday May 09 16:10 - 17:10 | Moscone Center - Hall E 133 |
Groovy jsr-223 script engine @ scripting.dev.java.net has been updated to use Groovy version 1.5.6.
If you want to learn more about Blu-ray disc and what Java has to do with it, you may want to attend the following talks/BOFs @ JavaOne 2008!
| Date/Time | Session ID | Session Name |
|---|---|---|
| Wednesday, May 07 9:30 AM - 10:30 AM | TS-5449 | Java™ Technology for Blu-ray™ and TV: Creating your own Blu-ray Java Discs |
| Wednesday, May 07 9:30 AM - 1:30 PM - 2:30 PM | TS-6464 | Blu-ray Disc Security |
| Wednesday, May 07 9:30 AM - 6:30 PM - 7:20 PM | BOF-5451 | Blu-ray and Java™ Technology Roundtable |
| Thursday May 08 1:30 PM - 2:30 PM | TS-5638 | Writing Connected Device Configuration Applications for Resource-Constrained Devices |
| Thursday May 08 1:30 PM - 2:30 PM | TS-5888 | Driving Innovation in Packaged Media (Blu-ray) User Experience |
From our group talk (TS-5449), we will be focusing on the open source project @ https://hdcookbook.dev.java.net. Meet you soon @ JavaOne !!
Are you interested in a byte code instrumentation (BCI) based dynamic tracing solution for the Java platform? If so, please visit https://btrace.dev.java.net. BTrace is a safe, dynamic tracing solution for Java. You can express tracing code in Java and run it against a running Java application. Your Java application should be running on JDK 6 or above for BTrace to work. You may be using VisualVM, the all-in-one Java troubleshooting tool. VisualVM supports plugin model to extend it's capabilities. BTrace plugin for VisualVM will be available soon. When BTrace plugin is available, you can trace your application from VisualVM tool. In the meanwhile, you can use BTrace command line tools. You may want to check out the user guide for the command line access.