feat(anywidget): Drop ipywidgets dependency #579
                
     Draft
            
            
          
  Add this suggestion to a batch that can be applied as a single commit.
  This suggestion is invalid because no changes were made to the code.
  Suggestions cannot be applied while the pull request is closed.
  Suggestions cannot be applied while viewing a subset of changes.
  Only one suggestion per line can be applied in a batch.
  Add this suggestion to a batch that can be applied as a single commit.
  Applying suggestions on deleted lines is not supported.
  You must change the existing code in this line in order to create a valid suggestion.
  Outdated suggestions cannot be applied.
  This suggestion has been applied or marked resolved.
  Suggestions cannot be applied from pending reviews.
  Suggestions cannot be applied on multi-line comments.
  Suggestions cannot be applied while the pull request is queued to merge.
  Suggestion cannot be applied right now. Please check back later.
  
    
  
    
Important
This is an experiment to tame the Python complexity of anywidget. We want to avoid breaking API changes for most anywidget projects; but I anticipate this will result in some breaking changes for widgets using low-level (undocumented) APIs.
The idea is to have anywidgets continue to subclass
anywidget.AnyWidgetin the same exact way, but cut outipywidgets.DOMWidgetfrom the inheritance chain (favoringtraitlets.HasTraitsinstead).This PR also allows us to dogfood the
anywidget.MimeBundleDescriptorwork. One challenge with having widgets subclass fromipywidgets.DOMWidgetis that all public (low-level) methods for creating widgets end up in public APIs to end users (i.e., poor encapsulation of comms).Rather than polluting the top-level Widget with comms-related public methods, all comms usage is confined to the
_repr_mimebundle_. This namespace is "private", and something that only anywidget authors would need to know about.Before
After
TODO:
FWIW, I think ideally widget authors would have more choice over the class for their widget (i.e.,
msgspec,pydantic,dataclass), but this is definitely a step in the right direction. If anything, it allows us to nail down APIs on the mimebundle descriptor that are "essential".