Switch jsmol to jupyter-jsmol

[mkoeppe]

Currently, jsmol is installed by the jmol spkg and then symlinked into the jupyter notebook directory.

In this ticket, we switch to ​https://pypi.org/project/jupyter-jsmol/ (​https://github.com/fekad/jupyter-jsmol), which can be installed by standard Python tools.

Discussion: ​https://groups.google.com/g/sage-devel/c/qKqTmLzHAbg/m/6YxQ_qWUBQAJ

See also:

• #30316 - Update jmol to 14.31.2, add spkg-configure for jmol
• #31027 - Make jmol optional

[mkoeppe]

(deleted)

[mkoeppe]

./sage -pip install jupyter-jsmol installs proper nbextensions:

$ jupyter nbextension list
Known nbextensions:
  config dir: /Users/mkoeppe/s/sage/sage-rebasing/local/etc/jupyter/nbconfig
    notebook section
      jupyter_jsmol/extension  enabled 
      - Validating: OK
      jupyter-js-widgets/extension  enabled 
      - Validating: OK

The installed files can be seen in $SAGE_LOCAL/share/jupyter/nbextensions/jupyter_jsmol/

[mkoeppe]

New commits:

​6617f87	build/pkgs/jupyter_jsmol: New
​cbaf661	build/pkgs/jmol/spkg-install.in: No longer install jsmol
​2ca5328	sage.repl.display.jsmol_iframe: Adjust path to jsmol
​ac1ab7c	sage.repl.ipython_kernel.install: No longer symlink jsmol

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​e6ebaf7

build/pkgs/jupyter_jsmol: Add dependencies, install-requires.txt

[mkoeppe]

sagetex declares jmol as a dependency - what is it used for?

[kcrisman]

sagetex declares jmol as a dependency - what is it used for?

I think for generating a 3d image in its documentation, compilation of which is part of its test suite.

[mkoeppe]

Hoping we can make progress on this ticket this week - ​https://wiki.sagemath.org/days111

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

​21259b2	build/pkgs/jupyter_jsmol: New
​b5f5f43	build/pkgs/jmol/spkg-install.in: No longer install jsmol
​a0fe543	sage.repl.display.jsmol_iframe: Adjust path to jsmol
​6e24bc0	sage.repl.ipython_kernel.install: No longer symlink jsmol
​52803b4	build/pkgs/jupyter_jsmol: Add dependencies, install-requires.txt
​7194459	build/pkgs/jmol/type: Make it optional
​0ed256c	build/make/Makefile.in: Compute SAGE_CHECK_... earlier, before evaluating dependencies
​1f714c1	build/pkgs/sagetex/dependencies: Conditionalize order-only deps on SAGE_CHECK_sagetex
​6a549d9	Merge branch 't/31020/build_make_makefile_in__fix_sage_check_logic__conditionalize_sagetex_dependencies_on_sage_check' into t/30315/switch_jsmol_to_jupyter_jsmol__make_jmol_optional

[mkoeppe]

Replying to kcrisman:

sagetex declares jmol as a dependency - what is it used for?

I think for generating a 3d image in its documentation, compilation of which is part of its test suite.

Thanks! I have adjusted the dependencies in #31020 (now a dependency) so that jmol is only treated as a dependency when SAGE_CHECK is active.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​3a45d72	build/pkgs/jupyter_jsmol/spkg-install.in: Do not rebuild the javascript file
​a8e0364	Specify bdist temp folder for WSL compatibility
​a3077c7	Use mktemp
​c5a14a3	build/bin/sage-dist-helpers (sdh_setup_bdist_wheel): New
​fb10429	build/pkgs/*/spkg-install.in: Use new function sdh_setup_bdist_wheel
​32bf463	Merge branch 't/31002/public/build/bdist_wsl' into t/30315/switch_jsmol_to_jupyter_jsmol__make_jmol_optional
​ae970c2	build/pkgs/jupyter_jsmol/spkg-install.in: Use new function sdh_setup_bdist_wheel

[kcrisman]

That is not a good idea. Because we want people to be able to build the documentation for SageTeX, which after all isn't "really" upstream. Instead, we should see if jsmol or some other tool suffices to build 3d pictures in SageTeX, which is a very useful use case for it. Untested = broken, and I don't think that we've ever had a dependency only in SAGE_CHECK before.

[kcrisman]

and I don't think that we've ever had a dependency only in SAGE_CHECK before.

Though my info on the build system is pretty outdated :-)

[mkoeppe]

Replying to kcrisman:

and I don't think that we've ever had a dependency only in SAGE_CHECK before.

Though my info on the build system is pretty outdated :-)

Actually, we've introduced a few of these this year because the complexity of Python test frameworks has escalated and we don't want to ship all of their dependencies. Currently, networkx, rpy2, and sage_sws2rst all have check-only dependencies.

[kcrisman]

Interesting. I wonder whether jsmol might be a drop-in replacement for this specific use case though. Or even Tachyon ...

[novoselt]

Replying to kcrisman:

Interesting. I wonder whether jsmol might be a drop-in replacement for this specific use case though. Or even Tachyon ...

What's the status of generating 3D plots and saving them using threejs? I think for SageTeX it is irrelevant what is producing pictures, just show that it works well. Tachyon is not the best in this regard.

[mkoeppe]

A discussion has unfolded in ​https://groups.google.com/g/sage-devel/c/qKqTmLzHAbg/m/6YxQ_qWUBQAJ

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

​e57f927	Merge branch 't/31002/public/build/bdist_wsl' into t/30315/switch_jsmol_to_jupyter_jsmol__make_jmol_optional
​017b816	build/pkgs/jupyter_jsmol: New
​db1b38c	build/pkgs/jmol/spkg-install.in: No longer install jsmol
​95a7131	sage.repl.display.jsmol_iframe: Adjust path to jsmol
​98b83f5	sage.repl.ipython_kernel.install: No longer symlink jsmol
​40a8823	build/pkgs/jupyter_jsmol: Add dependencies, install-requires.txt
​bf3c1dc	build/pkgs/jupyter_jsmol/spkg-install.in: Do not rebuild the javascript file

[mkoeppe]

I have narrowed this ticket to only separating the jmol and jsmol installation.

[mkoeppe]

Making jmol optional, and the discussion about new technologies to create images from 3d graphics, is now on #31027.

[gh-jcamp0x2a]

I was able to verify that JSmol still works correctly in both Jupyter and JupyterLab. However, I did notice a test failure in src/sage/repl/ipython_kernel/install.py due to the new deprecation warning:

File "src/sage/repl/ipython_kernel/install.py", line 142, in sage.repl.ipython_kernel.install.SageKernelSpec.use_local_jsmol
Failed example:
    spec.use_local_jsmol()
Expected nothing
Got:
    doctest:warning
      File "/home/joshua/sage/src/bin/sage-runtests", line 182, in <module>
        err = DC.run()
    [...]
    :
    DeprecationWarning: Symlinking jsmol is no longer necessary
    See https://trac.sagemath.org/30315 for details.

Also, after removing the local/share/jsmol folder since it is no longer needed (and wouldn't be present on a clean install), I get a couple more test failures in the same file:

File "src/sage/repl/ipython_kernel/install.py", line 144, in sage.repl.ipython_kernel.install.SageKernelSpec.use_local_jsmol
Failed example:
    os.path.isdir(jsmol)
Expected:
    True
Got:
    False

and

File "src/sage/repl/ipython_kernel/install.py", line 146, in sage.repl.ipython_kernel.install.SageKernelSpec.use_local_jsmol
Failed example:
    os.path.isfile(os.path.join(jsmol, "JSmol.min.js"))
Expected:
    True
Got:
    False

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​f2a02d9	Get rid of JSMOL_DIR
​7732e2e	build/pkgs/jmol/SPKG.rst: Update

[mkoeppe]

Thanks. This function, after all, does not make sense any more, so there's no point in trying to go through deprecation. I have now removed it completely (alongside with the variable JSMOL_DIR).

[gh-jcamp0x2a]

Looks good. JSmol still works in Jupyter and JupyterLab.

[mkoeppe]

Thank you!

[git]

Branch pushed to git repo; I updated commit sha1 and set ticket back to needs_review. New commits:

​80a3db1

build/pkgs/jupyter_jsmol/spkg-install.in: Use sdh_setup_bdist_wheel

[mkoeppe]

Forgot to actually use #31002 - this change is necessary for correct installation on WSL

[gh-jcamp0x2a]

Replying to mkoeppe:

Forgot to actually use #31002 - this change is necessary for correct installation on WSL

Ok. I tested this on Ubuntu WSL earlier without any installation issues. If I understand #31002 correctly, the problem was occurring due to mounting Windows folders, which I'm not doing, so that's probably why I didn't run into it.

I've gone ahead and re-verified that JSmol still works in Jupyter and JupyterLab.

[mkoeppe]

Thanks again!