Chet Haase's Blog

Chat with Chet and Scott

Another short dereference blog, in a series of short dereference blogs. This one's a pointer to the upcoming chat with Scott Violet and myself on Getting High Performance from Your Desktop Client.

The main message here is to log on to the chat site Tuesday, February 15th, at 11 am PST and send us some questions. The more people, the better the questions, the better the chat.

The secret message here, the reward for having read this far in this obscure blog entry, is that you have the opportunity to pre-stuff the ballot box. All chats start out slowly; everyone is just logging on, they're not sure whether to ask their question yet, a question may only occur to someone after they've seen other related chatter, whatever. So there are always a few questions that we have up our sleeve to start the conversation off (you know, things like "Scott, why is Swing just so darned wonderful?"). We can certainly come up with a few of these, but wouldn't it be better if you did instead?

Send some feedback into the blog below and we'll take these questions along to the chat with us. I can't promise they'll get asked; it depends on what the traffic is like. But I can promise that I'll see your questions below before I see any other questions during the chat...

Update: The chat transcript is now live; enjoy!

Comments

• Since I probably won't be online then, here goes...

  Is there a good way to directly take advantage of the Java 3D APIs for hardware acceleration from lightweight components? I'm looking at imbedding an interactive 3D scene within a document in a custom JEditorPane inside a JScrollPane. What hints would you have for squeezing out performance with an off-screen Buffer3D?

  William Tracy

  Posted by: afishionado on February 14, 2005 at 09:06 PM

• I haven't kept up with the Java2D integration with Swing and being able to offload Swing rendering of components to a GPU, but is this something we are able to utilize and is it something that would help speed screen drawing (and thus perceived performance).
  Michael

  Posted by: powerdroid on February 15, 2005 at 06:06 AM

• Also, any comments on how to objectively refute the rhetoric that comes from the SWT camp regarding Swings "performance problems". For those of us using Swing, usually there doesn't seem to be a problem...

  Posted by: powerdroid on February 15, 2005 at 07:06 AM

• I have to ask this everywhere about this bug related to Swing deployment
  http://www.sandrasf.com/other/sandra/files/sec.rtf and http://forums.java.net/jive/thread.jspa?threadID=315&tstart=0
  It's several years old, is there anyone who will address it?

  Also, what is the performance impact of creating threads, using syncronized blocks, etc.

  .V

  Posted by: netsql on February 15, 2005 at 08:09 AM

• I have a JTable with variable-height rows. Since SizeSequence requires O(N) to process inserts and removes, inserts and removes to my JTable are very slow.
  Is there any way to get the JTable to use an alternate SizeSequence?
  Would it be worthwhile to change SizeSequence to use a tree internally so that all operations are O(log N)?
  Cheers,Jessehttp://publicobject.com/glazedlists

  Posted by: jessewilson on February 15, 2005 at 08:41 AM

• I'd like to see some of your tought about handling of event in Swing.

  At the begining the event was overiding some method, then the listerner and anonymous class came.

  Now we have more opportunity like : java.beans.EventHandler, Swingworkers or Foxtrot.
  What is the approach you wil advice ?
  Do you have any reason for us not to use EvenHandler instead of writing tons of dummy anonymous class adapters ?

  (maybe a bit off topic,but anyway) Are you planning to improve swing with UI framework on top of swing that will be "action" based and configurable ? JDNC is fine, but it is not reducing much the maintenant pain on big J2SE applications. Some kind of struts or JSF thing for the client side ;-)
  (even more off topic, but please answer this) Anyplan to push javax.swing RI as a CDDL element (cf. OpenSolaris process) ? This will help much in people improving the Swing toolkit. Depending on the outcome of this process, it could validate/invalidate the idea of having the whole J2SE RI out under CDDL as well ;-)

  Thanks to all the swing team for having done the best toolkit API ever.

  
  PS: I do hope that the "upcoming" J2SE will bring some improvement on the datamodel which are not taking benefit of the collection framework and genericity to simplify and make transparent the data binding accross the various widgets
  

  Posted by: bjb on February 15, 2005 at 10:48 AM

• I have an HTML file that's about 9MB with 94,000 lines. It's basically a table with about 30,000 rows. Apart from the size, nothing fancy in the html. It takes about 17 seconds to load in IE. I tried to load it using JEditorPane and it takes ... I have no idea. Because after 10 minutes, I just killed the app.

  What is worse is that if I open another file, I would have expected JEditorPane to stop loading the previous file. But it doesn't. The parsing (or something) continues in the background hogging up the CPU. Is there a way to interrupt the page loading thread?

  I ended up just putting a message with a hyperlink in JEditorPane to load it using IE.

  Regards
  Milind

  Posted by: os2baba on February 19, 2005 at 04:55 AM

• 
  >It's basically a table with about 30,000 rows.

  I am afraid swing-html support was not designed to handle big html files.
  For example to handle one cell 10 objects are created at least.
  Say you have 5 cells per row than total number of cells is 150,000.
  1 500 000 objects is needed to handle cells only.
  Say average size of such object is 200 bytes.
  Than it takes about 300M to handle cells only.

  > The parsing (or something) continues in the background hogging up the CPU. Is there a way to interrupt the page loading thread?

  Loading thread is concealable (JEditorPane.setPage cancels current
  document loading) but it finishes itself at certain points only. It
  can quite only when accessing remote file. (See
  JEditorPane.PageStream class for details) At the time you cancel the
  loading swing-html might be stuck with those objects creation.
  Eventually loading will be canceled or OutOfMemory exception happens.

  (In fact you might monitor what happens with the VM using
  JConsole )

  
  If your data is table only then may I suggest as a workaround to use JTable?
  html file might be parsed using swing-html parser and injected into tableModel.
  Using customized tableModel it will need memory about the size of your
  html document which is at least 30 times less than what JEditorPane requires.
  

  Posted by: idk on February 22, 2005 at 01:18 PM

This work is licensed under a Creative Commons License.