support 3d graphics in sage mode in Jupyter notebooks

[williamstein]

Replicate:

• Open ipython notebook
• Choose current SageMath kernel
• In a codecell enter the code

x, y = var('x y')
S = sphere((0,0,0),size=0.3, color='red', aspect_ratio=[1,1,1])
show(S, figsize=8)

• No plot appears...

Correct url for jsmol.min.js is broken.

[williamstein]

This url at least has that file: http://chemapps.stolaf.edu/jmol/jsmol/

[williamstein]

Still broken:

https://cloud.sagemath.com/projects/4a5f0542-5873-4eed-a85c-a18c706e8bcd/files/support/2016-09-23-125146-issue159.html

[williamstein]

Bug since it silently fails.

[williamstein]

This is exactly this problem: jupyterhub/jupyterhub#401

It's I guess a bug in Sage. It's still a problem in sage-7.4

[williamstein]

If we make it so https://cloud.sagemath.com/nbextensions/jsmol/JSmol.min.js serves the file JSmol.min.js, then this may work fine for us. This JSmol.min.js is I think not updated much (maybe not once in 3 years). It's 223K and is in /projects/sage/sage-7.4/local/share/jsmol/JSmol.min.js.

FALSE: actually many more files are needed (maybe 20MB+ more)...

[williamstein]

No, that's just too ugly -- we really need to do this right by fixing sage. Also, our static content really should get served via /static.

[williamstein]

OK, giving up on this for now.

[williamstein]

I reported it upstream to the sage-devel list.

[williamstein]

Potentially very useful discussion: https://groups.google.com/forum/#!topic/sage-devel/VR6YEwCTsE4

[williamstein]

Maybe fixing this would just involve serving the relevant jmol 3d files from the right url via nginx. If you try to draw a 3d plot in jupyter sagemath mode, and watch the terminal, you'll see what absolute path from sage it fails to pull in. It's trying to download several megabytes of javascript/html/etc. -- some sort of automatic generated js (from java). It fails because Volker doesn't understand jupyter base url stuff. But we could actually fix it to work by just making that path be available (bundling it up as part of making the production site), and that would also be much faster (getting the html/js from our cdn rather than via an http tunnel).

[amirsagiv]

Hello, I just wanted to say that we've tested it on development version 7.5.beta3 (0c6824fff559e93b5517e53cd90ac1f1b04accf9) (git://github.com/sagemath/sage.git), and it seems to work. The test was performed on Debian testing xenial. All the examples were displayed correctly with Jupyter.
Thanks!

[williamstein]

@amirsagiv -- thanks. The problem is that Sage's 3d graphics are supported via some very hacky techniques, and do not properly support Jupyter's base_url functionality. This means they can't work in SMC, without some significant changes.

(For anybody reading this, 3d graphics do work well in Sage worksheets in SMC.)

[haraldschilly]

I've deployed a workaround, closing it.

[amirsagiv]

thanks. It works now!

…

On Fri, 25 Nov 2016 at 16:19 Harald Schilly ***@***.***> wrote: I've deployed a workaround, closing it. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#159 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AWmUCUY11dvO10gdAnBm2W1cDMAUbu2Lks5rBu6HgaJpZM4GMmWE> .