It was a good dicussion. The other panelists were James Velasco, Steven Reynolds, and Dan Sline. Dan started things off with a discussion of some proposed changes: closures, annotations for fields, etc.

I then did some demos. I did the BTrace, GChisto, and VisualVM demos that Jarda and I did during our technical session. The coolest demo I did, though, was Ken Russell's applet plugin demo. I dragged an applet out of the browser onto the desktop, closed the browser, and the applet kept running.

After that, Steven took over and talked about some of his favorite sessions, which included Gavin King's session on Web Beans and Mikael Grev's session on the MiG layout manager.

James wrapped things up with a neat approach: a JavaOne "index" that showed which trends were on their way up, and which were on their way down. Among those on the way up: NetBeans and JavaFX. Among those on their way down: SOAP-based web services.

Slides from the presenters are available here. A photo of the speakers is below. James and Dan are in front, that's Steven and me in the back row.

I did a technical session this year with Jaroslav Bachorík (who sometimes blogs here). Our goal was to do an overview of some of the tools available for tracking down performance and memory problems in Java applications. So the talk was broad, not deep - we covered 12 different tools and did demos of 3 of those.

It seemed to be well received, based on initial reviews. The room holds about 800 and was packed, so we did the session a second time late on Friday afternoon (not as large a crowd the second time, of course).

The slides are available here. There are several links to online resources, which I have reproduced below.

• Java SE Performance White Paper - a great starting point.


• Troubleshooting Guide for Java SE 6 with HotSpot VM - provides useful tips


• DTrace portal - contains links to several resources


• An article on using DTrace with Java applications, by Jarod Jenson.


• An article on using the Sun Studio Collector/Analyzer on Java applications, by Marty Itzkowitz.


• An interesting article by Nasser Nouri on using DTrace in conjunction with the Sun Studio Collector/Analyzer.


• The project page for BTrace.


• The project page for GChisto.


• Part 1 and part 2 of an article I wrote with Sundar and Frank Kieviet on dealing with different types of memory leaks in Java applications.


• Features in the NetBeans profiling tools, and the project page.


• The Hands On Lab on the NetBeans profiling tools that I taught with Maria Davydova is now available for download here.


• The project page for VisualVM.

This is sort of like asking if it would be possible to get another slice of grilled meat at a churrascaria; his answer was: "Of course!" :-)

In Brazil, everyone who knows Java knows Bruno - and even people who don't know Java know Bruno, including the president, as shown in this photo.

So Bruno said to me: "How busy do you want to be?" I responded: "Very." I'll think twice before I say that to him again.... :-D

I did presentations for the following Java users groups: SouJava, Java Noroeste, Java do Sertão, and ceJug. And I did a presentation at Universidade Católica de Brasília. The end result: within my first few days in the country, I traveled almost as many miles within Brazil as I had flown to get to Brazil.

But it was worth it. The food was excellent, the events were well-attended, and the best part (as usual) was getting to meet so many nice people. I was very well cared for in Brazil.

One of my goals was to show folks some of the new features coming in NetBeans IDE 6.1 (which is due out this week). One of the most important new features is powerful support for JavaScript. The support is built on what was put into the 6.0 release to support Ruby. As a result, powerful code-completion is included.

Now I'm not a JavaScript guru and have only played around with it a bit. Luckily, my co-worker Brian Leonard put together a very nice demo script that highlights the most important features. So I had been using that and had even managed to impress some hardcore JavaScript users with the editor's features at the first couple of events.

In Sertão, however, the code completion support did not work. Anyone who does as many live software demos as I do has been in this situation before - all the sudden, everything that worked great just moments before stops working:

I was stumped. Since it was the end of the final presentation of a long day, I went ahead and kept going and wrapped up my presentation without continuing that demo.

Immediately after I finished, one of the attendees rushed up to me and asked: "How can the editor provide JavaScript support outside of a <script> tag?" That's when it hit me like a ton of bricks: "Doh! I was adding JavaScript code into an .html file without a <script> tag. No wonder the editor did not work!"

So this is one of the few instances where one of my demos failed because of the presenter, instead of the software. In my defense, over the previous 80 hours I had only had about 12 hours of sleep, so I was a little off-balance. Nevertheless, the question made me realize what I was doing wrong, so I did the demo for a small group of folks who crowded around my laptop; the first feature of which is to show one of the flavors of code completion:

Photos from the Java Noroeste, Java do Sertão, and ceJug events are below.

They had noticed that I had talked about memory leaks to the Houston Java Users Group back in October and they asked me to do the same presentation for them.

So I drove up on March 12 and it was a great time. I got to chat briefly with Pete Carapetyan, who I had met last year; Pete has his own review of my presentation here. We had a good crowd: over 50 people. The slides are available here.

There were several really interesting comments and questions. One attendee commented on how the problem I demonstrated with a PermGen memory leak would not have happened if the application had not been repeatedly undeployed and redeployed to the web server. While that is certainly true, it led another attendee to comment: "Once I give them the .war file, I have no control over what they do with it." In other words, in an ideal world applications do not get repeatedly undeployed and redeployed, but in the real world things do not always work out that way....

There was also a question about cyclic object references. The Heap Walker in the NetBeans IDE provides annotations to indicate just those sorts of relationships. Cyclic references are shown with two green arrows that form a circle:

My thanks to Lisa Danzer for coordinating my visit to Dallas and (as always) to Aaron Houston for sending some books and t-shirts that we could use as prizes.

And many thanks to Erik Weibust of the MUG for inviting me up and running the meeting. I did not get to meet Erik last year because he was unable to attend that meeting. Turns out he is a great guy and for those of you in the area, he is one of the organizers of an interesting one-day conference: Dallas TechFest. It is on May 3 and the price cannot be beat: free!

I would consider driving up to attend, but I will be in California that day getting ready for NetBeans Day at JavaOne.

I will admit up front: the answer I provider here is a hack. As Brian Leonard points out in this IssueZilla entry, the IDE should have better support for this feature.

Let's say you are wondering whether or not Ctrl+Shift+Period is assigned to an action. Select Tools > Options > Keymap. You will see a list of the types of actions:

In order to find out whether Ctrl+Shift+Period is assigned, you need to attempt to assign it to an action. The easiest way to do this is to try to assign it to an action that does not currently have a keystroke assigned.

Expand the entry for Help and select the entry for About and click the Add button:

That brings up the Assign Shortcut dialog, where you can type in Ctrl+Shift+Period. If the keystroke combination you type in is already in use, the IDE will tell you to which action it is assigned. In my installation, that keystroke combination is already assigned:

Ken is the architect of the next-generation Java Plug-in and the project lead for JOGL, the Java binding to OpenGL.

Ken started his presentation with a couple of quotes, one from Ben Galbraith and another from the Java Posse, both of which can be summed up as: "Applets really suck." He then talked about the many reasons why, with a focus on problems in the Java browser plug-in.

In Ken's view, the newly rewritten Java browser plug-in changes everything. The problems with browsers hanging and crashing are gone. In addition, the deployment model for applets will be the same as for Web Start and the integration with JavaScript has improved. More details are in Ken's presentation slides, but just reading through them will not provide you with Ken's enthusiasm or the cool demos that he did.

The new browser plug-in is part of JDK 6 Update 10, which has also been referred to as the "Consumer JRE." It is also known as JDK 6 Update N, because originally they were not sure what update number to assign to it.

It all seemed to strike a chord with the audience - there were many questions. Rob Ratcliffe, one of the AJUG board members, wrote this in his recap email:

Ken gave a great talk on the new capabilities of the redesigned Java Plug-in. I'm excited that there may be a rebirth of embedded rich swing-based applications over the slightly unwieldy mix of DHTML, Javascript, Ajax, etc.. (Although, it appears that applets and Javascript will play well together in the new plug-in.)

Meanwhile, the OpenSolaris folks continue to make progress on creating a desktop-friendly version of Solaris. I recently got a nice new Sony laptop, so after I got Windows Vista and Ubuntu working on it I decided to try installing OpenSolaris.

Unfortunately, it would not install. I think the problem is this bug, but I am not certain. I tried the work around described in the bug, but it did not work for me. Bottom line was that even after getting considerable help from some very nice OpenSolaris gurus, I could not get the install to work with the partition that I had set aside on the laptop's hard disk.

I thought I was out of luck, and then came the announcement that Sun is planning to acquire Innotek (the deal has since been finalized). I had looked at VirtualBox briefly several months ago but I did not have an immediate need for it, so I had never installed it.

Once the announcement came out, several people started experimenting with it and blogging about it. So I tried it and it worked - I am typing this blog entry into an .html file using NetBeans IDE 6.0.1, running on OpenSolaris Developer Preview 2 (herein after referred to as OSDP2), which is running in a VirtualBox VM on Windows Vista.

Click the screen snapshot for a full-size version and you will see that NetBeans looks the same as it does on any other GNOME desktop (I do not know if OpenSolaris offers KDE or not...):

There were some rocks in the road. Read on for the full details....

I am using VirtualBox version 1.5.6. The first problem that I ran into was that OSDP2 would not install onto a fixed size VirtualBox virtual drive. I think that is because of a known issue in VirtualBox version 1.5.4, which is what I was using initially. The workaround is simple: I installed onto a dynamic virtual drive instead.

There were two significant problems after I installed: no network access and I could not set the video resolution of OSDP2 to anything higher than 1024x768. Alan Burlison has the solution for both problems in this blog entry. Note that by default, OSDP2 does not have an xorg.conf file - it does automatic configuration. So if you want to use the xorg.conf file that Alan provides, you will not be overwriting an existing xorg.conf - you will instead just be placing the file into /etc/X11.

Also note that in order to get the instruction that Alan provides for running the VBoxManage command to work, I had to not only shutdown the guest VM, but also the VirtualBox user interface itself.

Now comes the big question: Can I use this as my every day environment? I don't think so - not yet, anyway. The lag time as I type is very noticeable and using the mouse can be a bit frustrating. I am hoping those are just video/mouse driver problems. VirtualBox has guest operating system "Guest Additions" that help solve some of these issues. Unfortunately, there are no "Guest Additions" available for Solaris yet. Hopefully that will soon change.

In the meantime, it is good enough that I can use it to experiment with things such as the NetBeans DTrace plugin. :-)

Tor pointed out that the name "hippie completion" originated with the feature in XEmacs. Wherever the name came from, the concept is simple: in response to a control key combination, the editor fills in the nearest matching string. So for example, using NetBeans 6.0, on line 21 of this source file:

I have typed in myBi as the parameter to System.out.println(). If I then press Ctrl-K the editor will search backwards through the edit window for the nearest match and will then fill in the rest of the string for me:

Note that the match is done without any sort of context knowledge. That is an important point because it means that hippie completion is not smart, at least not in comparison to a standard code completion feature. The upside though is that it is faster than code completion.

If I had instead pressed Ctrl-Shift-K then the editor would have searched forward in the edit window and would have wrapped around to the beginning of the file and continued searching until it found myBig:

Some things to note:

• Press Ctrl-K (or Ctrl-Shift-K) repeatedly to have the editor suggest other strings that match.


• You don't have to type anything before invoking the feature - if you do not give it anything to match then the editor will just suggest the string that is closest to the cursor position.


• As mentioned, when it does its search the editor wraps around when it reaches the end of the file. It also wraps when it reaches the beginning. As a result, I always just use Ctrl-K, even when I want to insert a string that is further down in the file from where I am currently editing.


• The editor searches all open editor windows for a match, not just the window for the file that currently has focus.


• NetBeans has had this feature for years and years and prior to version 6.0 the keybindings were Ctrl-K to search backward and Ctrl-L to search forward. The addition of so many new features in NetBeans 6.0 required modifying the default keymap and the default forward search keybinding got changed from Ctrl-L to Ctrl-Shift-K.


• The official names for these features are "Insert Previous Matching Word" (Ctrl-K) and "Insert Next Matching Word" (Ctrl-Shift-K). But I prefer to use the term "Hidden Hippie." :-)

As always, there were many questions. Some I could answer, some I could not. For those that I could not answer, I did some asking around and Tor Norbye and Martin Krauskopf were kind enough to fill in the blanks for me.

Some of the highlights:

• What is the business model here? In other words, how is Sun going to make money on Ruby/Ruby on Rails development tools?
  We have a saying around here: "Volume drives value." Which means, in part, get the developer's attention up front and the money will come later. Specifically, at deployment time. I spoke with a number of attendees in Atlanta who are very interested in deploying Rails applications to GlassFish. And if they want a support contract on that runtime, Sun can provide that.


• What is the story on the lengthy scan times at startup?
  When the IDE creates a project it scans the source files and the libraries you have installed and caches some information. On a restart it will do a quick scan of your source to make sure the cache is still valid. But there are some problems in this area. I cannot reproduce the behavior under Windows or Linux, but others can. There is an IssueZilla (IZ) entry; see also this entry. And finally, some recent comments from Tor are worth reading.


• Is there support for HAML/SASS?
  Yes, but the last time I checked, it only works with NetBeans 6.0 - it does not work with development builds of NetBeans 6.1. More information here.


• Turning off the "Show Logical Project View" option is helpful, but not quite good enough because even with that option off, the vendor subdirectory of a Rails project is still not shown in the Projects window. Is there any plan to change that?
  As of December 7, in the 6.1 development builds the vendor subdirectory is shown in the Projects window.


• On large projects the code completion list is overwhelming - there is just too much stuff that shows up in the list. Is there any way to control this?
  The response I got from Tor was:

  The typical strategy for that is to type more before asking for completion. Don't try completion on empty context, or just one character. You do that when you really don't know what you're looking for - and in that case completion should probably show you what it knows about so you can go hunting.
  
  But if you type 3-4 chars it's no longer that long is it?
  
  Something I've been thinking of doing is to truncate searches at 50 matches. If there are more than 50 matches I will abort the search and insert a magic (unselectable) item in the completion list saying "Too many matches, result truncated." or something to that effect. That should make completion snappier and probably won't affect people too much since nobody really scrolls through 800 items looking for something anyway, you keep typing until what you're looking for shows up in the visible list.

  On a somewhat related note, there has recently been a discussion about this on one of the NetBeans Ruby email lists.


• If the vendor subdirectory for a project contains the Rails source then it gets indexed and that slows things down - is there any plan to fix this?
  Yes, that is fixed in the development builds, as Tor wrote: "... in 6.1 I've decoupled things a bit so that [the vendor subdirectory] is now shown in the projects view etc. but is still indexed differently than user sources."


• How does the IDE actually start the debugger?
  This is what I got from Martin:

  Instead of rdebug (standard Fast debugger for Ruby) we start rdebug-ide which is part of debug-commons project on RubyForge.
  
  [The] rdebug-ide script is executed with a another ruby script (debuggee) as a parameter. Currently only use case since rails is just another Ruby script.
  
  The rdebug-ide, in short:
  
  - starts the server to which a frontend connects and communicates through it
  - registers debug event hook in Ruby runtime
  - loads the debuggee (after 'start' command is received from the frontend)
  
  As events are happening they are sent through socket to the frontend and at the same time server is listening to the frontend (for commands like 'add breakpoint', 'do step', ...).

• Is there any way to get to a Ruby debug console while using the debugger?
  Currently, no. But Martin is aware of this request; there is an IZ entry for it.

I have never used DTrace in anger but have always wanted to learn more and Angelo delivered - I learned quite a bit. He covered the basic concepts and then drilled into numerous specific examples to show the value that DTrace provides. What was especially useful (and entertaining as well) were his war stories from the field. Those "real world" examples of DTrace being used to solve hard problems really made the topic lively.

As is normally the case, the evening started off with a short talk (a "technotizer" as it were :-) ). I was the speaker and I did an overview of just some of the new features in NetBeans IDE 6.0. I included a quick demo of some of the new editor features and then I turned it over to long-time NetBeans user (and AJUG board member) Rob Ratcliff, who did a quick demo of the new Swing Framework/Beans Binding features.

So all in all, a fun and informative evening. A few pictures below - for more check out Albert's photos here.

Once the support was added it was nice to be able to have things like code completion and some refactoring features, etc. Back in September I did a quick 10-minute demo of some of the Ruby support at the Lone Star Ruby Conference. I also met Keith Lancaster, the founder of the Houston Ruby and Rails User Group.

Keith invited me to do a presentation and that's how I ended up in Houston on a Tuesday night showing the Ruby and Rails features of the NetBeans IDE. It was a lot of fun - a very interactive session with many questions. Doing these sorts of presentations/demos is the best part of my job and one of the reasons I like it so much is that I end up learning new things along the way.

There were some questions that I could not answer, so after I got back home I got some answers from Tor Norbye, one of the developers of the Ruby support in NetBeans. Brian suggested I add the questions and answers to the NetBeans Ruby FAQ, which I have done. (Side note: the main NetBeans Ruby wiki page also has some great info. as does the main documentation page.)

• What's with the contents of ruby1/rubystubs?
  These stub methods are used by the IDE regardless of which Ruby interpreter you have configured. That's where I got confused - in my haste I misread the path which is why I had mistakenly thought these files were in the JRuby directory. More details here.


• Is there automated support for creating the self.down method of a Rails migration?
  Unfortunately, for now the answer is no.


• Can the IDE's generate dialog also run the Rails "delete" operations?
  Unfortunately, again the answer is no, but this will hopefully be fixed soon.


• Is there a way to turn off the "logical" Project view on Rails projects?
  This is a good example of where I learned something new. Most of the attendees would prefer that the IDE's Project tab display something flatter - more like the way its Files tab displays a Rails project. I did not realize that option was available, but it is. :-)


• Is there a default keyboard shortcut for the Rails generate dialog?
  This is a bit tricky.... If you right-click a Rails project the context menu includes an entry called "Generate" which brings up a dialog that provides easy access to the Rails generator scripts. There is no keyboard shortcut listed for that context menu entry. You can add one easily enough, however, it turns out that there is a default keyboard shortcut available when you are editing a .rb file in a Rails project.


• How can I run IRB with native Ruby?
  In NetBeans IDE 6.0, you cannot run the IRB shell with native Ruby - it always uses the JRuby that is bundled with the IDE. In newer daily builds, however, there is a workaround.

The answer to that last question about IRB brings up an interesting point. The feature set in NetBeans IDE 6.0 is frozen - the final bits will be available in just a few weeks. Work continues, however, on new features. Some of those new features will eventually be put on the beta update center (choose Tools > Plugins within the IDE to install plugins from an update center). But if you want to live on the edge, you can always install a nightly build. Or if you really want to be on the edge - install a build with just the Ruby pack from the continuous build server.

Jim put that list to a vote and the winner was my talk on tracking down memory leaks in Java applications. I originally did this presentation at JavaOne last year and have repeated it at JavaZone and TheServerSide Java Symposium.

It changes around a bit each time. The biggest addition was in July of this year for the TheServerSide conference in Barcelona: I added a whole section on permanent generation memory leaks, based on the work of Frank Kieviet and Edward Chou. Frank and Edward did an excellent presentation earlier this year at JavaOne dedicated entirely to that topic. Frank and I also collaborated with Sundar on an article about permanent generation memory leaks for Software Test & Performance magazine.

Another change that I made last night was to include a demo of VisualVM, a relatively new tool that is being developed as an open source project. There is an introductory document available on the project web site. I hope to write more about VisualVM soon.

The changes I have made to the presentation have also been driven by the questions that I have gotten when doing the presentation. A consistent theme that was repeated last night in Houston was: "Is there a way to prevent memory leaks from occurring in the first place?"

Automating detection is a problem. Tom Ball left a comment on that topic on my post from JavaZone. There are, however, some things that can be automated. If, for example, you are shadowing a static variable with a local variable and as a result you do not end up setting the static variable to null and a memory leak results, then a good editor or FindBugs can point out to you that you are shadowing the variable. But detecting that the shadowing indirectly leads to a memory leak is beyond the capability of any static analysis tool that I know of.

There are best practices that can certainly help. (For example, don't shadow static variables with local variables :-) ) I probably need to add a section on weak references to this presentation - I got a question specifically about that last night. My favorite post on that topic is from Ethan Nicholas: Understanding Weak References.

Some photos are below.

For a long time now, the NetBeans IDE has provided an easy way to move between the tabs: hold down the Ctrl key and then press the Tab key. You get a little popup that looks like this:

Continue to press the Tab key until the file you want is selected and then let go of the Ctrl key and the IDE will switch focus to that file's tab. Simple enough and very convenient.

One of the things that got some attention in the upcoming NetBeans IDE 6 release is keyboard short cuts, which is nice because I do not like to take my fingers off the keyboard. I stumbled across a couple of new short cuts earlier today: Ctrl-PageUp and Ctrl-PageDown. They are similar to the existing Ctrl-Tab but instead of displaying a list of the files, they just take switch focus to the next tab to the right (Ctrl-PageDown) or to the left (Ctrl-PageUp). One additional thing I like is that they wrap. In other words, pressing Ctrl-PageDown on the furthest tab to the right will set focus to the first tab on the left.

Update: A comment from grimlock81 indicates that this feature did exist in previous versions of NetBeans: Ctrl-Left arrow and Ctrl-Right arrow. In investigating further, it appears to me grimlock81 is close, but not quite correct. Using NetBeans 5.5, the default keystrokes are actually Alt-Left arrow and Alt-Right arrow.

So why didn't I spot that? Two reasons: this feature is not as easy to discover as I might like and the keymap entry moved in NetBeans 6. In NetBeans 5.5 the entries in the Options dialog's Keymap list for Previous Tab and Next Tab were under the "Go To" category. In NetBeans 6 they have been moved to the "Window/Select Document in" sub-category of the "Window" category:

One of these features in the editor is Mark Occurrences. It is pretty straightforward: put your cursor on something and the editor highlights all occurrences of that thing, be it a variable, method, method exit points, etc.

I should point out that this feature has actually been available for a long time now to NetBeans IDE users. The prolific Sandip Chitale wrote a plugin module for it long ago (I even created a screencast to demonstrate it). But in 6.0 the feature is included in the standard distribution.

Which is great. The only thing I'm not crazy about is that by default the highlighting marks stick around, even after I move the cursor away. The marks do not disappear until I put the cursor on something else that can be highlighted.

I would guess most folks prefer it this way, but I have always been a bit odd. :-) This is what user definable options are for - allowing each of us to do our own thing. In this case, Tools > Options provides:

The Options dialog not only allows me to control what sorts of things get highlighted, but I have also disabled the default "Keep Marks" option so that as soon as I move the cursor off of whatever was being highlighted, the highlights go away.

One final note: it appears that even though the control for Mark Occurrences is under the Java Code section in the Options dialog, the Ruby and JavaScript editors seem to abide by the setting of the "Keep Marks" option. I have filed an IssueZilla entry to get that improved....