Interact broken. SageMath 9.5 on Apple Mac OSX 12.1 . SageMath-9.5_x86_64

[Alan Stafford]

Interact just calls the function but no Interact widgets appear.

Example attached. It works at Cocalc.

[kcrisman]

Just for data points, which browsers did you try on this?  Thanks for reporting.

[Dima Pasechnik]

On Sat, Jan 29, 2022 at 6:28 PM kcrisman <kcri...@gmail.com> wrote:
>
> Just for data points, which browsers did you try on this? Thanks for reporting.

for what's worth, the interact works on Linux with Firefox.

>
> On Friday, January 28, 2022 at 1:02:57 PM UTC-5 alan_thoma...@yahoo.co.uk wrote:
>>
>> Interact just calls the function but no Interact widgets appear.
>>
>> Example attached. It works at Cocalc.
>

> --
> You received this message because you are subscribed to the Google Groups "sage-support" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-support...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-support/e0a15937-dcd4-49aa-9dce-cda82b3e3fa0n%40googlegroups.com.

[Alan Stafford]

I only used Safari. Safari supports widgets at Cocalc so I can't think that it won't support them should I run it locally. Interact has worked on previous versions of SageMath. I am hoping the Mac support team will look into this.

[Dima Pasechnik]

On Sat, Jan 29, 2022 at 8:05 PM 'Alan Stafford' via sage-support
<sage-s...@googlegroups.com> wrote:
>
> I only used Safari. Safari supports widgets at Cocalc so I can't think that it won't support them should I run it locally. Interact has worked on previous versions of SageMath. I am hoping the Mac support team will look into this.

Cocalc has its own implementation for handling ipython notebooks. No
wonder that it might be totally different locally, with different code
handling notebook.

Can you try with Chrome?

>
> On Saturday, January 29, 2022 at 6:40:10 PM UTC dim...@gmail.com wrote:
>>
>> On Sat, Jan 29, 2022 at 6:28 PM kcrisman <kcri...@gmail.com> wrote:
>> >
>> > Just for data points, which browsers did you try on this? Thanks for reporting.
>>
>> for what's worth, the interact works on Linux with Firefox.
>>
>> >
>> > On Friday, January 28, 2022 at 1:02:57 PM UTC-5 alan_thoma...@yahoo.co.uk wrote:
>> >>
>> >> Interact just calls the function but no Interact widgets appear.
>> >>
>> >> Example attached. It works at Cocalc.
>> >
>> > --
>> > You received this message because you are subscribed to the Google Groups "sage-support" group.
>> > To unsubscribe from this group and stop receiving emails from it, send an email to sage-support...@googlegroups.com.
>> > To view this discussion on the web visit https://groups.google.com/d/msgid/sage-support/e0a15937-dcd4-49aa-9dce-cda82b3e3fa0n%40googlegroups.com.
>
> --
> You received this message because you are subscribed to the Google Groups "sage-support" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-support...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-support/8f13034e-c737-4e4a-bc58-ceb6d7177281n%40googlegroups.com.

[Alan Stafford]

Exactly the same using Google Chrome.

[Alan Stafford]

I have found that it doesn't work if I launch from the SageMath 9.5 icon for any kernel 9.4 or 9.5. If I launch from the SageMath 9.4 interface it works for either kernel.

[Dima Pasechnik]

On Mon, Jan 31, 2022 at 3:59 AM 'Alan Stafford' via sage-support
<sage-s...@googlegroups.com> wrote:
>
> I have found that it doesn't work if I launch from the SageMath 9.5 icon for any kernel 9.4 or 9.5. If I launch from the SageMath 9.4 interface it works for either kernel.

I am not sure I follow here. How do you install Sage?

The basic thing that should work is starting Sage from terminal, as

sage -n

The macOS GUI we used to have is quite broken, and was removed, the
one we currently recommend is
https://github.com/3-manifolds/Sage_macOS/releases

>
>>>
>>>
> --
> You received this message because you are subscribed to the Google Groups "sage-support" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-support...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-support/7b98180b-e489-411b-a046-a8f30269d3f2n%40googlegroups.com.

[Alan Stafford]

I have downloaded both the SageMath 9.5 beta and 9.4 SageMath releases from git. I do not get widgets should I start the 9.4 or 9.5 kernel using the 9.5 interface, but I do when I use the 9.4 interface.

[Marc Culler]

It sounds to me like you do not have the SageMath 9.5 Jupyter kernel installed in /usr/local.  That gets installed when you run the Installer package that is included on the disk image.  Did you perhaps forget to run that installer after dragging the app to /Applications?

- Marc

[Alan Stafford]

I have run the Recommended_9_5.pkg file.

[Marc Culler]

Thanks.  And I can reproduce this.  I see the error shown below in the Firefox Console.  (I include it

here in case anyone might recognize it).

- Marc

Could not open comm  --  Error: Class jupyter.widget not found in registry   utils.js:930
    load_class                                           utils.js:894
    load_class                                           utils.js:879
    comm_open                                        comm.js:84
    i                                                          jQuery
    _handle_iopub_message                    kernel.js:1238
    _finish_ws_message                           kernel.js:1017
    _msg_queue                                        kernel.js:1008
    promise callback*Kernel.prototype._handle_ws_message kernel.js:1008
    i                                        jQuery
    reject                                utils.js:930
    (Async: promise callback)
    comm_open                    comm.js:85
    i                                       jQuery
    _handle_iopub_message kernel.js:1238
    _finish_ws_message       kernel.js:1017
    _msg_queue                   kernel.js:1008
    (Async: promise callback)
    _handle_ws_message    kernel.js:1008
    i                                     jQuery

[slelievre]

2022-02-01 15:13:27 UTC, Marc Culler:

>
> Thanks. And I can reproduce this. I see the error shown below

> in the Firefox Console. (I include it here in case anyone might

> recognize it).

There's an FAQ entry here

https://github.com/Wuliyuanulb/ComponentSDKDoc/blob/bfe53cf2162f62134356b58d7eae07e367044602/_build/_sources/faq.md.txt

that reads

> In browser F12 window, there is `Couldn't process kernel message`

> error in Console log. WrappedError message will be like:
> ```
> Error: Class jupyter.widget not found in registry at http://xxx
> ```
> This is because ipywidgets not correctly enabled in jupyter extension

> when using pip to install. Reinstall using conda will solve this issue.

> See [more details here](https://ipywidgets.readthedocs.io/en/latest/user_install.html).
> ```
> conda install -c conda-forge ipywidgets
> ```

And here are some more related discussions,

hoping there's something relevant to dig from them.

https://github.com/Caltech-IPAC/firefly_widgets/issues/6
https://github.com/InsightSoftwareConsortium/itkwidgets/issues/207
https://github.com/ipython-contrib/jupyter_contrib_nbextensions/issues/1319
https://github.com/jupyter-widgets/ipywidgets/issues/2569
https://github.com/jupyter-widgets/ipywidgets/issues/377
https://github.com/jupyterlab/jupyterlab/issues/4398
https://github.com/maartenbreddels/ipyvolume/issues/36
https://github.com/OpenGeoscience/geonotebook/pull/51
https://github.com/twosigma/beakerx/issues/6582

Found by searching using this request

https://github.com/search?q=Class+jupyter.widget+not+found+in+registry&type=issues

Not sure that helps at all, sorry.  --Samuel

[Marc Culler]

At some level the problem is that the Tornado web application used by Jupyter is returning a 404 when the client tries

to load:

http://localhost:8888/nbextensions/widgets/notebook/js/extension.js?v=20220201102631

I have no clue how the Tornado web application tries to translate that URL into a path to a static file named extension.js.

But that file does exist.  The difference between 9.4 and 9.5 with respect to that file is that it moved from local/share to

venv/share.  So I would guess that there is some part of how Jupyter configures its Tornado web application which

was not correctly adjusted to compensate for that change.

- Marc

[Marc Culler]

So is Sage following those instructions during its build process?  The docs do not say what those jupyter commands actually do.  They have separate instructions for enabling the widgets when using a virtualenv.  Of course Sage's venv is not a true virtualenv as created by venv, so it is hard to know whether the virtualenv instruction apply for Sage or not.  But given that the widgets worked before Sage switched to its virtualenv-like venv I would guess that they do apply.

It is unclear to me whether this is an issue with the packaging of the SageMath app for macOS or whether it is an issue with Sage's build process, which I follow when building Sage for the app.  It would be helpful to know the meaning of the sentence "Most of the time, installing ipywidgets automatically configures Jupyter Notebook to use widgets."

Specifically, what does it mean to "configure Jupyter Notebook to use widgets".

- Marc

[Matthias Koeppe]

Marc, are you referring to these instructions?

https://doc.sagemath.org/html/en/installation/launching.html#setting-up-sagemath-as-a-jupyter-kernel-in-an-existing-jupyter-notebook-or-jupyterlab-installation

[Matthias Koeppe]

On Tuesday, February 1, 2022 at 3:28:48 PM UTC-8 Marc Culler wrote:

Of course Sage's venv is not a true virtualenv as created by venv,

It is actually a true venv, unless no system python can be used and Sage builds its own copy of python3.

 

[Marc Culler]

Well, OK.  Maybe so.

But I've never seen a venv that resulted in a path like sage/local/var/lib/sage/venv-python3.9.9/var/lib/sage/.

- Marc

--
You received this message because you are subscribed to a topic in the Google Groups "sage-support" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/sage-support/FlLA_7tUb0M/unsubscribe.
To unsubscribe from this group and all its topics, send an email to sage-support...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sage-support/68d11539-11ba-4eac-aa91-db9926590d42n%40googlegroups.com.

[Marc Culler]

No.  I was referring to https://ipywidgets.readthedocs.io/en/latest/user_install.html which was Samuel's [more details here] link.

Incidentally, following those instructions did result in a log message indicating that the widgets had been enabled in Jupyter.  But there was no change in the behavior.  Attempting to load extension.js still produced a 404.

- Marc

--
You received this message because you are subscribed to a topic in the Google Groups "sage-support" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/sage-support/FlLA_7tUb0M/unsubscribe.
To unsubscribe from this group and all its topics, send an email to sage-support...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sage-support/26b228ed-ed16-400b-849c-7c3be030a413n%40googlegroups.com.

[Marc Culler]

What I said above appears not to be correct.  It looks like enabling the widgets extension with the instructions in the user_install page did work.  I am now getting a widget with Sage 9.5.  More testing is needed, but it looks promising.

- Marc

[Marc Culler]

Never mind.  I must have accidentally tested with 9.4.  Tornado is still returning a 404 when the notebook tries to load /nbextensions/widgets/notebook/js/extension.js.

[Marc Culler]

I figured it out.

* The root directory for the Tornado web application is sage/venv/share/jupyter (determined by trial and error).

* The path component in the url that the jupyter client uses to load the widget extension is

    /nbextensions/widgets/notebook/js/extension.js

* There is no directory named sage/venv/share/jupyter/nbextensions/widgets in the sage 9.5 tree.  Instead

there is a directory named sage/venv/share/jupyter/nbextensions/jupyter-js-widgets.

My fix was to create the directory sage/venv/share/jupyter/nbextensions/widgets/notebook

and a symlink sage/venv/share/jupyter/nbextensions/widgets/notebook/js -> ../../jupyter-js-widgets

With that change I get working widgets.  Experiment indicates that the jupyter-js-widgets directory

must exist.  But the widgets directory must also exist.  Why does the client use a path that does not

exist?  I do not know.  How can it be changed to use the correct path?  I was not able to figure that

out.  I could not find the code which the client uses to constructs its urls.  Maybe it is somewhere

inside a gigantic minimized javascript file.  Who knows?  Files containing only one line are not easily

searched with tools like grep.  I guess that is why facebook and twitter like them so much.

- Marc

[Matthias Koeppe]

On my machine, macOS with Sage 9.5 compiled from source, I see the 404s too, but the widget from the notebook work anyway. 

[Marc Culler]

That sounds very mysterious.  How can it create widgets if it cannot load the javascript file that implements the widgets?  Maybe the client finds the javascript file in some other path such as usr/local/jupyter and uses that file instead.  Of course a self-contained app cannot rely on anything like that.

- Marc

[Matthias Koeppe]

This looks like https://github.com/jupyter-widgets/ipywidgets/issues/1702 

[Marc Culler]

It certainly does.  But that ticket is from 2017. (???)

- Marc

[slelievre]

Pull request to fix Sage interacts on the CoCalc side:

- https://github.com/sagemathinc/cocalc/pull/5721

[Alan Stafford]

Interact is broken for SageMath 9.5, it works in 9.4 so I don't understand the pull request.

[Marc Culler]

They look like two unrelated problems.  In SageMath 9.5 there is a path problem which prevents the widgets nbextension.js file from being located by the Tornado server in some situations.  (That is fixed in the current release of the macOS app by creating a symlink at the path that the Tornado server uses for the nbextension.js file.)  The PR is about a possible bug in some version of jquery-ui.  The PR just updates to newer versions of jupyter-widgets, jquery and jquery-ui.

- Marc

--

You received this message because you are subscribed to a topic in the Google Groups "sage-support" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/sage-support/FlLA_7tUb0M/unsubscribe.
To unsubscribe from this group and all its topics, send an email to sage-support...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/sage-support/4b645611-eb44-485a-a1bf-6b2e18710534n%40googlegroups.com.

[kcrisman]

Just to follow up, did this issue get resolved satisfactorily in the 9.6 release cycle?  The last emails leave this unclear.

[Marc Culler]

This issue is resolved in the current release of the SageMath 9.5 macOS app.

- Marc

On Wed, Mar 30, 2022 at 1:39 PM kcrisman <kcri...@gmail.com> wrote:

Just to follow up, did this issue get resolved satisfactorily in the 9.6 release cycle?  The last emails leave this unclear.

--

You received this message because you are subscribed to a topic in the Google Groups "sage-support" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/sage-support/FlLA_7tUb0M/unsubscribe.
To unsubscribe from this group and all its topics, send an email to sage-support...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/sage-support/05f6596b-3f3f-49c5-8a10-0da585f9c553n%40googlegroups.com.

[kcrisman]

On Wednesday, March 30, 2022 at 2:48:15 PM UTC-4 Marc Culler wrote:

This issue is resolved in the current release of the SageMath 9.5 macOS app.

- Marc

Awesome.