Question/Poll on 3D plot behavior

140 views
Skip to first unread message

kcrisman

unread,
Jun 11, 2012, 3:36:24 PM6/11/12
to sage-...@googlegroups.com
At #12299 there is a really nice update to the Jmol interface which is being reviewed.  It makes it very easy to do (among other things)

Change colors on plots, use axes, etc. on the fly
Do things to just one or two of multiple plotted objects in a graphic (like showing mesh, color)
Export to something usable in a web page easily

And of course still has fun things like stereographic views, and it solves a number of memory issues and other things with the Java piece.

There are a few 'needs work' issues, but it sounds like Jonathan G. will be able to solve them.  One thing that is different in the interface, though, is due to making sure not too many are open at once (for memory), namely

Old: as soon as you evaluate, something goes live (rotatable etc.)
New: as soon as you evaluate, a static image is generated and you need to click for the interactive image to work

It sounds like there are some potential issues with going to an

In-Between: the first (whatever that means) image is "live" but subsequent images are not until you ask them to be

Given that any WebGL version will not be available for some time, we want to keep as many backends as possible, *and* that some of the control mentioned above would not be available in a first-gen (e.g.) three.js thing, I think it's worth the time to say whether the New is something you can live with.

Thanks,
- kcrisman

Volker Braun

unread,
Jun 11, 2012, 3:50:20 PM6/11/12
to sage-...@googlegroups.com
Even in the current implementation you need to press a mouse button before you can drag the view around. So I don't quite understand what is different in the new jmol. Can you elaborate on that?

What should be avoided is that you first have to press&release mouse button before you can press again & drag. At least not without any visual feedback that the first mouse click changed any state. If, say, some arrows appear then it would be fine imho.

kcrisman

unread,
Jun 11, 2012, 4:47:33 PM6/11/12
to sage-...@googlegroups.com


On Monday, June 11, 2012 3:50:20 PM UTC-4, Volker Braun wrote:
Even in the current implementation you need to press a mouse button before you can drag the view around. So I don't quite understand what is different in the new jmol. Can you elaborate on that?


In the new implementation, there is a static image at the start - not the case before.  To get a "live" image, one has to click on a button which says something like "Sleeping - make interactive" or something, wait for the Jmol to load up, and then start dragging etc.

 
What should be avoided is that you first have to press&release mouse button before you can press again & drag. At least not without any visual feedback that the first mouse click changed any state. If, say, some arrows appear then it would be fine imho.


Much more than arrows - I've attached some screenshots.
Screen shot 2012-06-11 at 4.44.13 PM.png
Screen shot 2012-06-11 at 4.44.45 PM.png
Screen shot 2012-06-11 at 4.44.57 PM.png
Screen shot 2012-06-11 at 4.45.28 PM.png

Eviatar

unread,
Jun 11, 2012, 8:16:07 PM6/11/12
to sage-...@googlegroups.com
Looks much better! I have no problems with it requiring the extra click, assuming it generates the still image fairly quickly. Is the viewing angle for the still customizable? Will it still behave the old way in the terminal?

P Purkayastha

unread,
Jun 12, 2012, 3:34:37 AM6/12/12
to sage-...@googlegroups.com


On Tuesday, June 12, 2012 4:47:33 AM UTC+8, kcrisman wrote:


On Monday, June 11, 2012 3:50:20 PM UTC-4, Volker Braun wrote:
Even in the current implementation you need to press a mouse button before you can drag the view around. So I don't quite understand what is different in the new jmol. Can you elaborate on that?


In the new implementation, there is a static image at the start - not the case before.  To get a "live" image, one has to click on a button which says something like "Sleeping - make interactive" or something, wait for the Jmol to load up, and then start dragging etc.

The new implementation definitely deserves a +1. On a local server it is not so obvious, but loading jmol applets takes a long time when running it remotely. This also means that the webpage is loading and if there are too many applets then it also freezes the browser (especially a bit older browsers) for a while.

Loading static images is faster and more useful in these scenarios. The extra click is not a bother since the page will load fast in the beginning and the overall experience is better.

Volker Braun

unread,
Jun 12, 2012, 6:37:39 AM6/12/12
to sage-...@googlegroups.com
On Tuesday, June 12, 2012 8:34:37 AM UTC+1, P Purkayastha wrote:
The new implementation definitely deserves a +1. On a local server it is not so obvious, but loading jmol applets takes a long time when running it remotely.

Who is generating the static image, the Sage server or the client-side browser? In the first case, does that mean that Java is now a prerequisite for running the server?

kcrisman

unread,
Jun 12, 2012, 9:05:15 AM6/12/12
to sage-...@googlegroups.com, gu...@uwosh.edu
I think it's the same Jmol program either way, but I don't know where it lives, since I'm just testing on my local machine.  Presumably the same answer to your question as before this, since if you try to do plot3d without Java it doesn't work, I guess (unless you use a different viewer).

Jonathan Gutow

unread,
Jun 12, 2012, 9:16:50 AM6/12/12
to sage-devel
The static image is generated server side, using Jmol if a JVM is installed on the server or Tachyon if the JVM is unavailable.  The Jmol image is more complete, so we should recommend that a JVM be installed on the server,but it is not necessary.

Note, that if you are on a device that does not have Java, you now at least get an static image.  My next move is to replace the static image with the option for a spinning .gif.  However, let's get this in.

Jonathan
                         Dr. Jonathan H. Gutow
Chemistry Department                                gu...@uwosh.edu
UW-Oshkosh                                          Office: 920-424-1326
800 Algoma Boulevard                                FAX:920-424-2042
Oshkosh, WI 54901
                http://www.uwosh.edu/facstaff/gutow

Volker Braun

unread,
Jun 12, 2012, 9:19:56 AM6/12/12
to sage-...@googlegroups.com
Sounds good to me!

Jonathan Gutow

unread,
Jun 12, 2012, 9:29:23 AM6/12/12
to sage-devel
Note, I'm running a test public server at 


You can create an account and try it.  Please note this is running on a virtual machine on my home computer, it is not very hefty.  They've been playing with the firewall on campus and I cannot use my somewhat heftier work server because I have other things running on the ports they are allowing through the firewall.

Send feedback.  If you find a bug please put it on trac #12299 (http://trac.sagemath.org/sage_trac/ticket/12299)

Thanks,
Jonathan

kcrisman

unread,
Jun 12, 2012, 10:15:57 AM6/12/12
to sage-...@googlegroups.com


On Monday, June 11, 2012 8:16:07 PM UTC-4, Eviatar wrote:
Looks much better! I have no problems with it requiring the extra click, assuming it generates the still image fairly quickly. Is the viewing angle for the still customizable? Will it still behave the old way in the terminal?



Yes, the static image comes up pretty quickly.  As far as I can tell the command-line behavior is the same as it was before.  I know nothing about viewing angles :)

kcrisman

unread,
Jun 12, 2012, 10:31:38 AM6/12/12
to sage-...@googlegroups.com


On Monday, June 11, 2012 8:16:07 PM UTC-4, Eviatar wrote:
Looks much better! I have no problems with it requiring the extra click, assuming it generates the still image fairly quickly. Is the viewing angle for the still customizable? Will it still behave the old way in the terminal?


Eviatar and ppurka, if you're using Linux, could you try out the test server linked at #12299 and just report how things work with FF and/or other browsers on it?  (I suppose the FF clones would be useful, since they sometimes have Java issues, though I don't know if it's worth trying to fix.)

Anyone who has access to Windows would have feedback that is especially welcome.  Just make an account, open a worksheet, and do a few 3d plots using the latest IE or Chrome or something.  Thanks!

P Purkayastha

unread,
Jun 12, 2012, 11:00:39 AM6/12/12
to sage-...@googlegroups.com


On Tuesday, June 12, 2012 10:31:38 PM UTC+8, kcrisman wrote:


On Monday, June 11, 2012 8:16:07 PM UTC-4, Eviatar wrote:
Looks much better! I have no problems with it requiring the extra click, assuming it generates the still image fairly quickly. Is the viewing angle for the still customizable? Will it still behave the old way in the terminal?


Eviatar and ppurka, if you're using Linux, could you try out the test server linked at #12299 and just report how things work with FF and/or other browsers on it?  (I suppose the FF clones would be useful, since they sometimes have Java issues, though I don't know if it's worth trying to fix.)


The plots work on firefox-10 but doesn't work on the current stable or unstable version of opera. this is on 64bit gentoo linux. I will some more tests later tomorrow on a ubuntu 12.04 (64bit).

I would also say that high quality should be turned on by default, the plots look so much better. I also shared a worksheet with you (kcrisman) which had several 3d plots that you can test (we used it here a couple of months ago).

kcrisman

unread,
Jun 12, 2012, 11:20:20 AM6/12/12
to sage-...@googlegroups.com

I would also say that high quality should be turned on by default, the plots look so much better. I also shared a worksheet with you (kcrisman) which had several 3d plots that you can test (we used it here a couple of months ago).

I agree about the high quality, and have already mentioned that on the ticket, though I assume there is some technical reason not to do so. 

Jonathan

unread,
Jun 12, 2012, 12:01:56 PM6/12/12
to sage-...@googlegroups.com

On Tuesday, June 12, 2012 10:20:20 AM UTC-5, kcrisman wrote:

I would also say that high quality should be turned on by default, the plots look so much better. I also shared a worksheet with you (kcrisman) which had several 3d plots that you can test (we used it here a couple of months ago).

I agree about the high quality, and have already mentioned that on the ticket, though I assume there is some technical reason not to do so. 
We could try going to high quality as the default.  I originally did this to save memory.  Many of the issues with Jmol were related to available applet memory in the browser environment.  Since I am now limiting the number of open applets, this should be less of an issue.  I will try switching it.  I will let you know when that option is available on my test server. 

Jason Grout

unread,
Jun 12, 2012, 1:42:49 PM6/12/12
to sage-...@googlegroups.com
Was it the high quality (i.e., antialiasing) that caused all the white
speckles to appear all the time?

Jason



Jonathan

unread,
Jun 12, 2012, 4:52:37 PM6/12/12
to sage-...@googlegroups.com
I have set up my test server to default to high quality (antialiasing).  Test it out.  

The answer to Jason's question is "yes".  The antialiasing causes white speckles in regions of high contrast change.  That was another reason to start with it off.  The easiest way to see this is to plot a surface.  Turn on the black mess and start it spinning.  Then toggle the "high quality" on and off.

Jonathan

Eviatar

unread,
Jun 12, 2012, 7:48:31 PM6/12/12
to sage-...@googlegroups.com
The image generation is working great on Google Chrome on Linux, with IcedTea. However, the color and mesh options don't all work. Only some of the colours do (the lighter ones, for some reason), and the mesh and translucency options don't seem to do anything except causing crashes.


On Monday, 11 June 2012 12:36:24 UTC-7, kcrisman wrote:

Jonathan

unread,
Jun 12, 2012, 8:24:42 PM6/12/12
to sage-...@googlegroups.com
Maybe, IcedTea is improving...Are you using release 7?  I cannot get many Java applets (not just Jmol) to launch under the present "stable" release (6.x).  I have resorted to installing Oracle/Sun Java on my Linux machines.

I agree that I have had better luck with the applications under IceTea.  At least one of my test servers uses IcedTea and the static images generated using the headless Java application does work on that server.  I have just not had much luck with applets embedded in browsers.

The other way to read your message is that you never get the applet to launch in the browsers.  Is that the case?  If so, that is my experience.  See http://www.webupd8.org/2012/01/install-oracle-java-jdk-7-in-ubuntu-via.html for what looks like an easy way to install the Oracle Java under Ubuntu.

Jonathan

Jonathan

unread,
Jun 12, 2012, 8:26:43 PM6/12/12
to sage-...@googlegroups.com


Just had a power outage.  My test server will be down until after I get a chance to eat dinner....sorry

Jonathan

Eviatar

unread,
Jun 12, 2012, 8:31:27 PM6/12/12
to sage-...@googlegroups.com
The applet does work on IcedTea 6, but some of the mesh and color options don't, as I mentioned. But yes, I'll try out Java 6 and see how that goes.

Can you set the plot to any colour on your end? Because I thought it was strange that it only works with light colours.

Jonathan

unread,
Jun 12, 2012, 9:24:09 PM6/12/12
to sage-...@googlegroups.com
Test server at http://gutow.no-ip.org:8080 is up again.

On Tuesday, June 12, 2012 7:31:27 PM UTC-5, Eviatar wrote:
The applet does work on IcedTea 6, but some of the mesh and color options don't, as I mentioned. But yes, I'll try out Java 6 and see how that goes.
I may have to try it again.  IcedTea is definitely improving if your experience is any indication. 

Can you set the plot to any colour on your end? Because I thought it was strange that it only works with light colours.

Yes I can select any of the colors in the little color picker. 

P Purkayastha

unread,
Jun 12, 2012, 11:54:29 PM6/12/12
to sage-...@googlegroups.com

On Tuesday, June 12, 2012 11:00:39 PM UTC+8, P Purkayastha wrote:


The plots work on firefox-10 but doesn't work on the current stable or unstable version of opera. this is on 64bit gentoo linux. I will some more tests later tomorrow on a ubuntu 12.04 (64bit).


I tested on ubuntu-12.04 64bit. It works on chrome but not in opera.

Jonathan

unread,
Jun 13, 2012, 9:26:52 AM6/13/12
to sage-...@googlegroups.com
I can reproduce the problem with Opera on MacOS as well.  The error messages all relate to jqueryui css stuff.  So I'm not sure what is going on.  I do use some of that, but I'm just using the Sage defaults...I will investigate, but am not optimistic.

Jonathan 

Jonathan

unread,
Jun 13, 2012, 11:20:24 AM6/13/12
to sage-...@googlegroups.com
Testing on Win 7:

IE7: Static image displays.  Clicking on the make interactive button gives the warning dialog about "many" of the advanced features not working and that the user should change browsers.  Then nothing happens.  The javascript that launches the applet seems to be ignored.  It fails completely silently, with no errors in the console.  I don't know how to begin tracing this problem.  I think it is probably a total loss for using the applet in IE at this point.

FF: Seems to work fine.

Can somebody test Chrome on Win 7?

Jonathan

P Purkayastha

unread,
Jun 13, 2012, 11:41:45 AM6/13/12
to sage-...@googlegroups.com


On Wednesday, June 13, 2012 9:24:09 AM UTC+8, Jonathan wrote:
Test server at http://gutow.no-ip.org:8080 is up again.

On Tuesday, June 12, 2012 7:31:27 PM UTC-5, Eviatar wrote:
The applet does work on IcedTea 6, but some of the mesh and color options don't, as I mentioned. But yes, I'll try out Java 6 and see how that goes.
I may have to try it again.  IcedTea is definitely improving if your experience is any indication. 

This is not good for linux distributions because many of the linux distributions have defaulted to IcedTea for java and that's what will be installed by default. Going forward,  it will start needing more manual setup to get oracle/sun java installed (that means no official support in most cases).

Jonathan

unread,
Jun 13, 2012, 12:30:47 PM6/13/12
to sage-...@googlegroups.com
I agree this is a problem for Linux users.  I know that the IcedTea project is aware of the problems with loading many applets and Javascript <-> applet communication using the IceTea plugin. From reading the bugs list, it appears they may be testing some fixes in version 7.  I haven't yet set up a machine with the beta IcedTea to see if that fixes the problems.  I'm keeping my fingers crossed that IcedTea will start working better.  Prior versions would not even run the Jmol application.  Now Jmol runs and the applet loads and has some communication with Javascript.

From my perspective as a Java application developer, the Oracle Java is the official version and we work hard to maintain compatibility with that.  It is up to the IcedTea people to have IcedTea be equivalent.

Jonathan

William Stein

unread,
Jun 13, 2012, 12:40:43 PM6/13/12
to sage-...@googlegroups.com
I'm glad Javascript developers don't have that perspective...

>
> Jonathan
>
> --
> To post to this group, send an email to sage-...@googlegroups.com
> To unsubscribe from this group, send an email to
> sage-devel+...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/sage-devel
> URL: http://www.sagemath.org



--
William Stein
Professor of Mathematics
University of Washington
http://wstein.org

kcrisman

unread,
Jun 13, 2012, 1:54:24 PM6/13/12
to sage-...@googlegroups.com


On Wednesday, June 13, 2012 11:20:24 AM UTC-4, Jonathan wrote:
Testing on Win 7:

IE7: Static image displays.  Clicking on the make interactive button gives the warning dialog about "many" of the advanced features not working and that the user should change browsers.  Then nothing happens.  The javascript that launches the applet seems to be ignored.  It fails completely silently, with no errors in the console.  I don't know how to begin tracing this problem.  I think it is probably a total loss for using the applet in IE at this point.


Or at least older IEs.  You mean IE 9 wouldn't work either?  IE 10 is apparently around the corner.

Jonathan Gutow

unread,
Jun 13, 2012, 3:55:11 PM6/13/12
to sage-...@googlegroups.com

On Jun 13, 2012, at 10:20 AM, Jonathan wrote:

> Testing on Win 7:
>
> IE7: Static image displays. Clicking on the make interactive button gives the warning dialog about "many" of the advanced features not working and that
^^^^
OOPS! I meant IE9. Previous tests on IE7 and 8 give some functionality. Just not all of the things work (e.g. toggling the advanced controls off for example). This is because IE7/8 seemed to handle jqueryui toggle functions differently depending on where they are on the page. I never could figure out why.

Jonathan

Jonathan Gutow

unread,
Jun 13, 2012, 4:09:57 PM6/13/12
to sage-...@googlegroups.com

On Jun 13, 2012, at 11:40 AM, William Stein wrote:


From my perspective as a Java application developer, the Oracle Java is the
official version and we work hard to maintain compatibility with that.  It
is up to the IcedTea people to have IcedTea be equivalent.

I'm glad Javascript developers don't have that perspective...
William,

I resent that statement and think it shows a misunderstanding about what is not working...The problem with the IcedTea browser plugin is associated with mishandling of data passing by the plugin.  This means things are mangled or in many cases simply not passed completely from the browser to the applet.  Another way of saying this is that IcedTea does not implement its APIs as specified (they still have major bugs to fix). The  only way I would have to fix that is to rewrite the plugin code.  There is a limit to how many projects one can take on, plus I'm probably not the best person to work on it.  Basically, to make things work in IcedTea we would simply have to give up on the functions that don't work  in it.  It seems better to build things to work in the standard Java since the goal of IcedTea is to make it operationally equivalent, which they are fast approaching.  My feeling is that the Linux community is jumping over to IcedTea as a standard prematurely.  That said users have a relatively straightforward option for converting back to the standard Java if they need to (see my previous comments).

You may have noticed that I have the same kinds of problems with IE which often does not abide by HTML or ECMAScript (JavaScript) standards.  This does not mean I don't try to make things work in IE, but at some point you have to decide if the effort is worth it.  I have and do spend a lot of time figuring out ways to make things work in as many environments as possible, but do not have infinite time.

Jonathan

kcrisman

unread,
Jun 14, 2012, 1:34:11 PM6/14/12
to sage-...@googlegroups.com


On Tuesday, June 12, 2012 4:52:37 PM UTC-4, Jonathan wrote:
I have set up my test server to default to high quality (antialiasing).  Test it out.  

The answer to Jason's question is "yes".  The antialiasing causes white speckles in regions of high contrast change.  That was another reason to start with it off.  The easiest way to see this is to plot a surface.  Turn on the black mess and start it spinning.  Then toggle the "high quality" on and off.



Wow, it's sort of mesmerizing.

I'm a little torn.  The speckles are definitely distracting, if of potential value for hypnosis.  But the high quality looks enough better that maybe it's worth having them.  Is there a way to have the mesh turn-on button automatically make the quality lower again?  I'm only seeing this with mesh, I assume this is the only time it happens.

Jonathan

unread,
Jun 14, 2012, 3:16:19 PM6/14/12
to sage-...@googlegroups.com


On Thursday, June 14, 2012 12:34:11 PM UTC-5, kcrisman wrote:
surface.  Turn on the black mess and start it spinning.  Then toggle the "high quality" on and off.



Wow, it's sort of mesmerizing.

I'm a little torn.  The speckles are definitely distracting, if of potential value for hypnosis.  

Fixed by defaulting to very dark blue instead of black for default mess.  Thank you Karl-Dieter for the idea.  Included in the lastest patch on trac #12299.

You can still get the speckles by choosing a black mesh manually ;)

Jonathan 

Jonathan

unread,
Jun 16, 2012, 4:54:51 PM6/16/12
to sage-...@googlegroups.com
Trac #12299 is in need of a careful final review.  I think I've caught everything.  Thanks for the feedback.

Jonathan
Reply all
Reply to author
Forward
0 new messages