Multi-range sliders for DataFilterExtension (#340)

Closes #339 

### Change list

- Add `MultiRangeSlider`, a subclass of `ipywidgets.VBox` to connect
multiple float range sliders
- Add to API docs

docs/api/controls.md

@@ -0,0 +1,3 @@
+# lonboard.controls
+
+::: lonboard.controls.MultiRangeSlider

lonboard/__init__.py

@@ -1,7 +1,7 @@
 """Python library for fast, interactive geospatial vector data visualization in Jupyter.
 """
 
-from . import colormap, traits
+from . import colormap, controls, traits
 from ._layer import (
     BaseArrowLayer,
     BaseLayer,

lonboard/controls.py

@@ -0,0 +1,81 @@
+from functools import partial
+from typing import Sequence
+
+import traitlets
+from ipywidgets import FloatRangeSlider
+from ipywidgets.widgets.trait_types import TypedTuple
+
+# Import from source to allow mkdocstrings to link to base class
+from ipywidgets.widgets.widget_box import VBox
+
+
+class MultiRangeSlider(VBox):
+    """A widget for multiple ranged sliders.
+
+    This is designed to be used with the
+    [DataFilterExtension][lonboard.experimental.DataFilterExtension] when you want to
+    filter on 2 to 4 columns on the same time.
+
+    If you have only a single filter, use an ipywidgets
+    [FloatRangeSlider][ipywidgets.widgets.widget_float.FloatRangeSlider] directly.
+
+    # Example
+
+    ```py
+    from ipywidgets import FloatRangeSlider
+
+    slider1 = FloatRangeSlider(
+        value=(2, 5),
+        min=0,
+        max=10,
+        step=0.1,
+        description="First slider: "
+    )
+    slider2 = FloatRangeSlider(
+        value=(30, 40),
+        min=0,
+        max=50,
+        step=1,
+        description="Second slider: "
+    )
+    multi_slider = MultiRangeSlider([slider1, slider2])
+    multi_slider
+    ```
+
+    Then to propagate updates to a rendered layer, call
+    [jsdlink][ipywidgets.widgets.widget_link.jsdlink] to connect the two widgets.
+
+    ```py
+    from ipywidgets import jsdlink
+
+    jsdlink(
+        (multi_slider, "value"),
+        (layer, "filter_range")
+    )
+    ```
+
+    As you change the slider, the `filter_range` value on the layer class should be
+    updated.
+    """
+
+    # We use a tuple to force reassignment to update the list
+    # This is because list mutations do not get propagated as events
+    # https://github.com/jupyter-widgets/ipywidgets/blob/b2531796d414b0970f18050d6819d932417b9953/python/ipywidgets/ipywidgets/widgets/widget_box.py#L52-L54
+    value = TypedTuple(trait=TypedTuple(trait=traitlets.Float())).tag(sync=True)
+
+    def __init__(self, children: Sequence[FloatRangeSlider], **kwargs):
+        # We manage a list of lists to match what deck.gl expects for the
+        # DataFilterExtension
+        def callback(change, *, i: int):
+            value = list(self.value)
+            value[i] = change["new"]
+            self.set_trait("value", value)
+            self.send_state("value")
+
+        initial_values = []
+        for i, child in enumerate(children):
+            func = partial(callback, i=i)
+            child.observe(func, "value")
+            initial_values.append(child.value)
+
+        super().__init__(children, value=initial_values, **kwargs)

lonboard/experimental/layer_extension.py

@@ -148,6 +148,68 @@ class DataFilterExtension(BaseExtension):
     )
     ```
 
+    The `DataFilterExtension` allows filtering on 1 to 4 attributes at the same time. So
+    if you have four numeric columns of interest, you can filter on the intersection of
+    all of them.
+
+    For easy visualization, we suggest connecting the `DataFilterExtension` to an
+    interactive slider from `ipywidgets`.
+
+    ```py
+    from ipywidgets import FloatRangeSlider
+
+    slider = FloatRangeSlider(
+        value=(2, 5),
+        min=0,
+        max=10,
+        step=0.1,
+        description="Slider: "
+    )
+    slider
+
+    jsdlink(
+        (slider, "value"),
+        (layer, "filter_range")
+    )
+    ```
+
+    If you have 2 to 4 columns, use a
+    [`MultiRangeSlider`][lonboard.controls.MultiRangeSlider], which combines multiple
+    `FloatRangeSlider` objects in a form that the `DataFilterExtension` expects.
+
+    ```py
+    from ipywidgets import FloatRangeSlider, jsdlink
+
+    slider1 = FloatRangeSlider(
+        value=(2, 5),
+        min=0,
+        max=10,
+        step=0.1,
+        description="First slider: "
+    )
+    slider2 = FloatRangeSlider(
+        value=(30, 40),
+        min=0,
+        max=50,
+        step=1,
+        description="Second slider: "
+    )
+    multi_slider = MultiRangeSlider([slider1, slider2])
+    multi_slider
+
+    jsdlink(
+        (multi_slider, "value"),
+        (layer, "filter_range")
+    )
+    ```
+
+    # Important notes
+
+    - The DataFilterExtension only supports float32 data, so integer data will be casted
+      to float32.
+    - The DataFilterExtension copies all data referenced by `get_filter_value` to the
+      GPU, so it will increase memory pressure on the GPU.
+
     # Layer Properties
 
     ## `filter_enabled`

mkdocs.yml

@@ -44,6 +44,7 @@ nav:
           - api/layers/base-layer.md
       - api/basemap.md
       - api/colormap.md
+      - api/controls.md
       - api/traits.md
       - Experimental:
           - Layer Extensions:
@@ -144,6 +145,7 @@ plugins:
             - https://geoarrow.github.io/geoarrow-rs/python/latest/objects.inv
             - https://geopandas.org/en/stable/objects.inv
             - https://geopy.readthedocs.io/en/stable/objects.inv
+            - https://ipywidgets.readthedocs.io/en/stable/objects.inv
             - https://matplotlib.org/stable/objects.inv
             - https://numpy.org/doc/stable/objects.inv
             - https://pandas.pydata.org/pandas-docs/stable/objects.inv

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "lonboard"
-version = "0.5.0"
+version = "0.6.0-beta.1"
 description = "Python library for fast, interactive geospatial vector data visualization in Jupyter."
 authors = ["Kyle Barron <kyle@developmentseed.org>"]
 license = "MIT"