Revert to reactive property approach. Only provide decorators for packages that need coercion

Author: cpsievert

CHANGELOG.md

@@ -7,6 +7,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [UNRELEASED]
 
+* The `@render_widget` decorator now attaches a `widget` (and `value`) attribute to the function it decorates. This allows for easier access to the widget instance (or value), and eliminates the need for `register_widget` (which is now soft deprecated).  (#119)
 * Reduce default plot margins on plotly graphs.
 
 ## [0.2.4] - 2023-11-20

shinywidgets/__init__.py

@@ -10,7 +10,6 @@
 from ._render_widget import (
     render_altair,
     render_bokeh,
-    render_leaflet,
     render_plotly,
     render_pydeck,
     render_widget,
@@ -22,7 +21,6 @@
     "render_widget",
     "render_altair",
     "render_bokeh",
-    "render_leaflet",
     "render_plotly",
     "render_pydeck",
     # Reactive read second

shinywidgets/_as_widget.py

@@ -35,7 +35,7 @@ def as_widget(x: object) -> Widget:
 
 def as_widget_altair(x: object) -> Optional[Widget]:
     try:
-        from altair import JupyterChart
+        from altair import JupyterChart  # pyright: ignore[reportMissingTypeStubs]
     except ImportError:
         raise RuntimeError(
             "Failed to import altair.JupyterChart (do you need to pip install -U altair?)"
@@ -55,9 +55,9 @@ def as_widget_bokeh(x: object) -> Optional[Widget]:
     # TODO: ideally we'd do this in set_layout_defaults() but doing
     # `BokehModel(x)._model.sizing_mode = "stretch_both"`
     # there, but that doesn't seem to work??
-    from bokeh.plotting import figure
+    from bokeh.plotting import figure  # pyright: ignore[reportMissingTypeStubs]
 
-    if isinstance(x, figure):
+    if isinstance(x, figure):  # type: ignore
         x.sizing_mode = "stretch_both"  # pyright: ignore[reportGeneralTypeIssues]
 
     return BokehModel(x)  # type: ignore
@@ -67,7 +67,7 @@ def as_widget_plotly(x: object) -> Optional[Widget]:
     # Don't need a try import here since this won't be called unless x is a plotly object
     import plotly.graph_objects as go  # pyright: ignore[reportMissingTypeStubs]
 
-    if not isinstance(x, go.Figure):
+    if not isinstance(x, go.Figure):  # type: ignore
         raise TypeError(
             f"Don't know how to coerce {x} into a plotly.graph_objects.FigureWidget object."
         )

shinywidgets/_render_widget.py

@@ -2,50 +2,41 @@
 
 from typing import TYPE_CHECKING
 
-from ipywidgets.widgets import Widget  # pyright: ignore[reportMissingTypeStubs]
-
 if TYPE_CHECKING:
-    from altair import JupyterChart
+    from altair import JupyterChart  # pyright: ignore[reportMissingTypeStubs]
     from jupyter_bokeh import BokehModel  # pyright: ignore[reportMissingTypeStubs]
     from plotly.graph_objects import (  # pyright: ignore[reportMissingTypeStubs]
         FigureWidget,
     )
     from pydeck.widget import DeckGLWidget  # pyright: ignore[reportMissingTypeStubs]
-
-    # Leaflet Widget class is the same as a Widget
-    # from ipyleaflet import Widget as LeafletWidget
+else:
+    JupyterChart = BokehModel = FigureWidget = DeckGLWidget = object
 
 from ._render_widget_base import ValueT, WidgetT, render_widget_base
 
 __all__ = (
     "render_widget",
     "render_altair",
     "render_bokeh",
-    "render_leaflet",
     "render_plotly",
     "render_pydeck",
 )
 
-
-class render_widget(render_widget_base[ValueT, Widget]):
+# In the generic case, just relay whatever the user's return type is
+# since we're not doing any coercion
+class render_widget(render_widget_base[WidgetT, WidgetT]):
     ...
 
-
+# Package specific renderers that require coercion (via as_widget())
+# NOTE: the types on these classes should mirror what as_widget() does
 class render_altair(render_widget_base[ValueT, JupyterChart]):
     ...
 
-
 class render_bokeh(render_widget_base[ValueT, BokehModel]):
     ...
 
-
-class render_leaflet(render_widget_base[WidgetT, WidgetT]):
-    ...
-
-
 class render_plotly(render_widget_base[ValueT, FigureWidget]):
     ...
 
-
 class render_pydeck(render_widget_base[ValueT, DeckGLWidget]):
     ...