Fix registry state lookup #3653

bollwyvl · 2022-12-18T19:25:43Z

References

fixes issue introduced on fix: make .widget and .widget_types deprecated #3567

Changes

use _registry.get instead of register.get
update official actions/* to latest versions to squash deprecation warnings
split up and upload failing results of Test: JavaScript job

Motivation

This is relatively difficult to demonstrate, as it requires a widget that creates a new client-side widget. For a reproducer see this in-development notebook:

run all cells
try to add a new .ipynb

File "/lib/python3.10/site-packages/ipykernel/comm.py", line 201, in comm_open
    f(comm, msg)
  File "/lib/python3.10/site-packages/ipywidgets/widgets/widget.py", line 421, in handle_comm_opened
    widget_class = register.get(state['_model_module'],
AttributeError: 'function' object has no attribute 'get'

it's possible to update the file with this patch, by downloading it and replacing /lib/python3.10/site-packages/ipywidgets/widgets/widget.py

Future Work

Likely this should probably just get reviewed, and hopefully merged, as it's fairly clearly an improvement by inspection.

A test would emulate a client creating a new, pre-registered Widget.

However, a better long-term check would be applying mypy, pyflakes or the new-fangled (but very fast) ruff or some related tool which would do a better job at statically analyzing the code for catching these kinds of issues, but is probably out of scope for this one-line fix.

github-actions · 2022-12-18T19:25:56Z

👈 Launch a binder notebook on branch bollwyvl/ipywidgets/fix-register-is-a-function

maartenbreddels · 2022-12-19T14:19:24Z

Thanks @bollwyvl !
I've pushed a test that will avoid regressions.

And I agree, this could have been avoided. I'll put it on the agenda for next time we have a widget meeting.
(cc @jasongrout who was interested in this as well)

bollwyvl · 2022-12-19T15:00:31Z

python/ipywidgets/ipywidgets/widgets/tests/test_widget.py

@@ -74,3 +78,27 @@ def test_compatibility():
    caller_path = inspect.stack(context=0)[1].filename
    assert all(x.filename == caller_path for x in record)
    assert len(record) == 6
+
+
+def test_create_from_frontend():


bollwyvl · 2022-12-19T15:04:59Z

Beyond static analysis: it's probably worth pulling off the band-aid and applying black, isort, etc. Heck, I've even been using ssort recently on some projects... just more highly subjective things not to have to make a decision/talk about on PRs (after talking about the initial PR). I think just doing it, with a single git history ignore thing, rather than the slow darken process IPython has been following, is easier for most involved.

maartenbreddels · 2022-12-19T15:18:29Z

Yes, I think we all totally agree with you. I'd love to see it. What's your opinion on using pre-commit?

maartenbreddels · 2022-12-19T15:19:34Z

assert 'application/vnd.jupyter.widget-view+json' in mime_bundle, "widget should have have a view"

Also had this error locally...

jasongrout · 2022-12-19T16:02:21Z

it's probably worth pulling off the band-aid and applying black, isort

FYI, #3566 moves to precommit with all those things applied.

bollwyvl · 2022-12-19T16:40:16Z

Fire, pre-commit is great... in CI. If it makes PRs to fix stuff it finds. I don't let it touch my machine, too many unmanaged environments. 🤣

maartenbreddels · 2022-12-20T18:06:53Z

Hmm, some CI rot? Two seemingly unrelated workflows are failing. Is this something you touched in this PR @bollwyvl, or do you happen to know how to fix this?

bollwyvl · 2022-12-21T02:30:40Z

seemingly unrelated workflows

Well, they both fail around playwright, of which I can't claim a lot of knowledge... I'm more concerned about the persist OoM on RTD, and the lack of meaningful output on the karma tests.

Without some of these things pinned (or at least archived) during a build, it's hard to say what has shifted. Perhaps a kicked run of main would shed some light on whether this is localized to this PR... I can roll back a bunch of changes, as I don't think there's much doubt about the critical line of the initial commit on this PR.

bollwyvl · 2022-12-21T03:27:54Z

I've slimmed this back down to just the fix and the tests, with de23fa6 being the last commit with the CI changes.

bollwyvl · 2022-12-21T03:29:00Z

python/ipywidgets/ipywidgets/widgets/tests/test_send_state.py

@@ -12,29 +12,20 @@
 # A widget with simple traits
 class SimpleWidget(Widget):
    a = Bool().tag(sync=True)
-    b = Tuple(Bool(), Bool(), Bool(), default_value=(False, False, False)).tag(sync=True)
+    b = Tuple(Bool(), Bool(), Bool(), default_value=(False, False, False)).tag(


can i roll these back, too?

bollwyvl · 2022-12-21T04:15:14Z

Even though the error is being discovered by playwright, i'll wager the issue lies with jupyter_server 2.0.2... haven't had a chance to look through the changes:

jupyter-server/jupyter_server@v2.0.1...v2.0.2

bollwyvl · 2022-12-21T04:16:18Z

Maybe this:

jupyter-server/jupyter_server@v2.0.1...v2.0.2#diff-72888b4a95043904b226c1aa5addc9687a1dd09706da5e59ed28be58ae5f99c9R2820

bollwyvl · 2022-12-21T04:42:43Z

jupyter-server/jupyter_server#1143

bollwyvl · 2022-12-21T05:03:55Z

jupyter-server/jupyter_server#1144

bollwyvl · 2022-12-21T13:17:52Z

That was merged... maybe a release forthcoming? Then we can get back to the other CI problems here, i suppose...

bollwyvl · 2022-12-21T14:18:14Z

ok, that cleared up those two emergent fails. i'm still not sure if blocking this on fixing the docs/js CI fails is the right play, and would prefer to start a second PR to investigate further... though this work did lead to finding two upstream issues, so at least there's that.

maartenbreddels · 2022-12-21T14:53:09Z

Really appreciate the help here Nick!
I'm ok with merging this with docs failing (although that also needs fixing). If we can get the javascript failure sorted, that would make a lot of people happy (and I could do a new release).

bollwyvl · 2022-12-21T20:03:39Z

Ok, tried a lot of things, but turns out it was just ubuntu-latest vs ubuntu-20.04. I rolled back everything except that change, and it appears to be passing.

I've removed the firefox install, as i don't know (or really care) what karma needs to be happy, but it's likely if it does need other things like geckodriver, that they might not be properly updated when updating firefox. It can be hard to reason about the compatibility windows of these things. On a side note: i've been using firefox and geckodriver from conda-forge for some years now, and they've only rarely lead me astray: we try to keep both the latest and the latest LTS available. Firefox in particular is a binary repack, so can cause issues (like missing x11) on e.g. lightweight containers, but github runners have 50gb of crap already installed, so it hasn't been an issue. Also, they aren't snaps (buh, gross).

maartenbreddels · 2022-12-21T21:28:12Z

Awesome work. I'm happy to merge this from this PR. I think I'll rebase/squash it into 2 or 3 commits somewhere tomorrow. Or, if you have your own opinion on how many commits there should be, feel free to do it yourself (I'm ok with force pushing)

bollwyvl · 2022-12-21T22:26:21Z

Squash it all!

maartenbreddels · 2022-12-22T08:33:08Z

I don't question authority 🙇 😄

maartenbreddels · 2022-12-22T09:25:34Z

This is released in ipywidgets 8.0.4

bollwyvl · 2022-12-22T14:48:58Z

Thanks all! I can take a look at the docs as well, but might get a little tinker-y with it.

Thoughts on going to pydata-sphinx-theme and myst-nb? I have basically given up trying to get nbsphinx to keep running.

maartenbreddels · 2022-12-22T14:59:03Z

I have basically given up trying to get nbsphinx to keep running.

Why is that?

I haven't followed that development much, but I think you are saying myst-nb should replace nbsphinx?

I kind of like the pydata-sphinx-theme, looks clean. Maybe opening an issue on this so we can get some opinions in?

register is a function

43b0b38

This was referenced Dec 18, 2022

fix: make .widget and .widget_types deprecated #3567

Merged

Test / JavaScript build failing on all PRs #3651

Open

bollwyvl added 4 commits December 18, 2022 13:44

some debugging of js tests

b9e40f9

remove extra node_modules cache

cbcaf4f

bump action versions for deprecation warnings

9db5f7c

more step naming

cd03fca

bollwyvl force-pushed the fix-register-is-a-function branch from 40bfd60 to 23eccbc Compare December 18, 2022 20:58

try some docs pins

5ef6573

bollwyvl force-pushed the fix-register-is-a-function branch from 23eccbc to 5ef6573 Compare December 18, 2022 21:00

test: create widget from frontend

e16d5a8

bollwyvl commented Dec 19, 2022

View reviewed changes

bollwyvl added 2 commits December 19, 2022 09:01

remove ipykernel pin

4c9eda1

remove ipykernel pin (conda)

8fcd447

fix test, setup and teardown messed things up, refactor

de23fa6

bollwyvl added 2 commits December 20, 2022 21:25

roll back ci changes to simplify ci

651ac23

merge upstream

cf85800

bollwyvl commented Dec 21, 2022

View reviewed changes

bollwyvl closed this Dec 21, 2022

bollwyvl reopened this Dec 21, 2022

bollwyvl force-pushed the fix-register-is-a-function branch from cd39116 to 3a12432 Compare December 21, 2022 19:49

use ubuntu-20.04 for js CI

3ac0ab3

bollwyvl force-pushed the fix-register-is-a-function branch from 3a12432 to 3ac0ab3 Compare December 21, 2022 19:50

maartenbreddels merged commit eacd961 into jupyter-widgets:master Dec 22, 2022

bollwyvl mentioned this pull request Dec 22, 2022

Overhaul docs build chain #3656

Closed

bollwyvl deleted the fix-register-is-a-function branch December 22, 2022 15:07

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix registry state lookup #3653

Fix registry state lookup #3653

bollwyvl commented Dec 18, 2022 •

edited

Loading

github-actions bot commented Dec 18, 2022

maartenbreddels commented Dec 19, 2022

bollwyvl Dec 19, 2022

bollwyvl commented Dec 19, 2022

maartenbreddels commented Dec 19, 2022

maartenbreddels commented Dec 19, 2022

jasongrout commented Dec 19, 2022

bollwyvl commented Dec 19, 2022

maartenbreddels commented Dec 20, 2022

bollwyvl commented Dec 21, 2022

bollwyvl commented Dec 21, 2022

bollwyvl Dec 21, 2022

bollwyvl commented Dec 21, 2022

bollwyvl commented Dec 21, 2022

bollwyvl commented Dec 21, 2022

bollwyvl commented Dec 21, 2022

bollwyvl commented Dec 21, 2022

bollwyvl commented Dec 21, 2022

maartenbreddels commented Dec 21, 2022

bollwyvl commented Dec 21, 2022

maartenbreddels commented Dec 21, 2022

bollwyvl commented Dec 21, 2022 via email

maartenbreddels commented Dec 22, 2022

maartenbreddels commented Dec 22, 2022

bollwyvl commented Dec 22, 2022

maartenbreddels commented Dec 22, 2022

Fix registry state lookup #3653

Fix registry state lookup #3653

Conversation

bollwyvl commented Dec 18, 2022 • edited Loading

References

Changes

Motivation

Future Work

github-actions bot commented Dec 18, 2022

maartenbreddels commented Dec 19, 2022

bollwyvl Dec 19, 2022

Choose a reason for hiding this comment

bollwyvl commented Dec 19, 2022

maartenbreddels commented Dec 19, 2022

maartenbreddels commented Dec 19, 2022

jasongrout commented Dec 19, 2022

bollwyvl commented Dec 19, 2022

maartenbreddels commented Dec 20, 2022

bollwyvl commented Dec 21, 2022

bollwyvl commented Dec 21, 2022

bollwyvl Dec 21, 2022

Choose a reason for hiding this comment

bollwyvl commented Dec 21, 2022

bollwyvl commented Dec 21, 2022

bollwyvl commented Dec 21, 2022

bollwyvl commented Dec 21, 2022

bollwyvl commented Dec 21, 2022

bollwyvl commented Dec 21, 2022

maartenbreddels commented Dec 21, 2022

bollwyvl commented Dec 21, 2022

maartenbreddels commented Dec 21, 2022

bollwyvl commented Dec 21, 2022 via email

maartenbreddels commented Dec 22, 2022

maartenbreddels commented Dec 22, 2022

bollwyvl commented Dec 22, 2022

maartenbreddels commented Dec 22, 2022

bollwyvl commented Dec 18, 2022 •

edited

Loading