[REVIEW] from_geopandas converts GeoPandas GeoSeries objects into cuspatial GeoArrow form.

CHANGELOG.md

@@ -60,7 +60,7 @@ Please see https://github.com/rapidsai/cuspatial/releases/tag/v0.20.0a for the l
 - PR #320 Fix quadtree construction bug: zero out `device_uvector` before `scatter`
 - PR #328 Fix point in polygon test for cudf::gather breaking change
 
-# cuSpatial 0.16.0 (Date TBD)
+# cuSpatial 0.16.0 (21 Oct 2020)
 
 ## New Features
 - PR #290 Add Java bindings and a cuSpatial JAR package for Java users.

conda/environments/cuspatial_dev_cuda11.0.yml

@@ -10,7 +10,7 @@ dependencies:
   - cudf=0.20.*
   - cudatoolkit=11.0
   - gdal>=3.0.2
-  - geopandas=0.7.0
+  - geopandas>=0.9.0
   - cmake>=3.18
   - cython>=0.29,<0.30
   - gtest

conda/environments/cuspatial_dev_cuda11.1.yml

@@ -10,7 +10,7 @@ dependencies:
   - cudf=0.20.*
   - cudatoolkit=11.1
   - gdal>=3.0.2
-  - geopandas=0.7.0
+  - geopandas>=0.9.0
   - cmake>=3.14
   - cython>=0.29,<0.30
   - gtest

conda/environments/cuspatial_dev_cuda11.2.yml

@@ -10,7 +10,7 @@ dependencies:
   - cudf=0.20.*
   - cudatoolkit=11.2
   - gdal>=3.0.2
-  - geopandas=0.7.0
+  - geopandas>=0.9.0
   - cmake>=3.14
   - cython>=0.29,<0.30
   - gtest

conda/recipes/cuspatial/meta.yaml

@@ -30,6 +30,7 @@ requirements:
     - cudf {{ minor_version }}.*
     - rmm {{ minor_version }}.*
     - gdal >=3.2.0,<3.3.0a0
+    - geopandas >=0.9.0,<0.10.0a0
 
 test:
   commands:

docs/source/api.rst

@@ -1,62 +1,84 @@
-~~~~~~~~~~~~~~~~~~~~~~~
-cuSpatial API Reference
-~~~~~~~~~~~~~~~~~~~~~~~
+GeoPandas Compatibility
+-----------------------
 
+We support any geometry format supported by `GeoPandas`. Load geometry information from a `GeoPandas.GeoSeries` or `GeoPandas.GeoDataFrame`.
 
-GIS
----
-.. currentmodule:: cuspatial
+    >>> gpdf = geopandas.read_file('arbitrary.txt')
+        cugpdf = cuspatial.from_geopandas(gpdf)
 
-.. autofunction:: cuspatial.directed_hausdorff_distance
-.. autofunction:: cuspatial.haversine_distance
-.. autofunction:: cuspatial.lonlat_to_cartesian
-.. autofunction:: cuspatial.point_in_polygon
-.. autofunction:: cuspatial.polygon_bounding_boxes
-.. autofunction:: cuspatial.polyline_bounding_boxes
+or
 
+    >>> cugpdf = cuspatial.GeoDataFrame(gpdf)
 
-Indexing
---------
 .. currentmodule:: cuspatial
 
-.. autofunction:: cuspatial.quadtree_on_points
+.. autoclass:: cuspatial.GeoDataFrame
+        :members:
+        :show-inheritance:
+.. autoclass:: cuspatial.GeoSeries
+        :members:
+        :show-inheritance:
+.. autoclass:: cuspatial.geometry.geocolumn.GeoColumn
+        :members:
+        :show-inheritance:
+.. autoclass:: cuspatial.GeoArrowBuffers
+        :members:
+        :show-inheritance:
+
+
+Spatial Indexing
+--------
 
+Spatial indexing functions provide blisteringly-fast on-GPU point-in-polygon
+operations.
 
-Interpolation
--------------
 .. currentmodule:: cuspatial
 
-.. autoclass:: CubicSpline
-.. automethod:: CubicSpline.__init__
-.. automethod:: CubicSpline.__call__
-
+.. autofunction:: cuspatial.quadtree_point_in_polygon
+.. autofunction:: cuspatial.quadtree_point_to_nearest_polyline
+.. autofunction:: cuspatial.point_in_polygon
+.. autofunction:: cuspatial.polygon_bounding_boxes
+.. autofunction:: cuspatial.polyline_bounding_boxes
+.. autofunction:: cuspatial.quadtree_on_points
+.. autofunction:: cuspatial.join_quadtree_and_bounding_boxes
+.. autofunction:: cuspatial.points_in_spatial_window
 
-Spatial Querying
-----------------
-.. currentmodule:: cuspatial
 
-.. autofunction:: cuspatial.points_in_spatial_window
+GIS
+---
 
+Two GIS functions make it easier to compute distances with geographic coordinates.
 
-Spatial Joining
-----------------
 .. currentmodule:: cuspatial
 
-.. autofunction:: cuspatial.join_quadtree_and_bounding_boxes
-.. autofunction:: cuspatial.quadtree_point_in_polygon
-.. autofunction:: cuspatial.quadtree_point_to_nearest_polyline
+.. autofunction:: cuspatial.haversine_distance
+.. autofunction:: cuspatial.lonlat_to_cartesian
+
 
 Trajectory
 ----------
+
+Trajectory functions make it easy to identify and group trajectories from point data.
+
 .. currentmodule:: cuspatial
 
 .. autofunction:: cuspatial.derive_trajectories
-.. autofunction:: cuspatial.trajectory_bounding_boxes
 .. autofunction:: cuspatial.trajectory_distances_and_speeds
+.. autofunction:: cuspatial.directed_hausdorff_distance
+.. autofunction:: cuspatial.trajectory_bounding_boxes
+.. autoclass:: CubicSpline
+.. automethod:: CubicSpline.__init__
+.. automethod:: CubicSpline.__call__
 
 
 IO
 --
+
+cuSpatial offers native GPU-accelerated shapefile reading. In addition, any host-side GeoPandas DataFrame can be copied into GPU memory for use with cuSpatial
+algorithms.
+
 .. currentmodule:: cuspatial
 
 .. autofunction:: cuspatial.read_polygon_shapefile
+.. autofunction:: cuspatial.from_geopandas
+

docs/source/conf.py

@@ -21,7 +21,6 @@
 #
 import os
 import sys
-import cudf
 
 sys.path.insert(0, os.path.abspath("../.."))
 
@@ -35,6 +34,7 @@
 # extensions coming with Sphinx (named 'sphinx.ext.*') or your custom
 # ones.
 extensions = [
+    "sphinx.ext.autosectionlabel",
     "sphinx.ext.intersphinx",
     "sphinx.ext.autodoc",
     "sphinx.ext.autosummary",
@@ -197,5 +197,6 @@
 numpydoc_class_members_toctree = False
 nbsphinx_allow_errors = True
 
+
 def setup(app):
     app.add_stylesheet("params.css")

docs/source/index.rst

@@ -1,12 +1,36 @@
 Welcome to cuSpatial's documentation!
 =====================================
 
+cuSpatial provides functionality for fast GPU-based spatial index and join
+functions via point-in-polygon, a pathing library for trajectory identification
+and reconstruction, and accelerated GIS functions like haversine distance and
+grid projection. The icing-on-the-cake of cuSpatial is that we integrate neatly
+with `GeoPandas` and RAPIDS `cudf`.
+
+This enables you to take advantage of all of the features of `GeoPandas`, plus
+blisteringly fast GPU acceleration for pandas-like functionality in `cudf` and
+spatial functions that perform quickly on millions and tens of millions of
+geometries.
+
+GeoArrow
+--------
+
+cuSpatial proposes a new GeoArrow format from the fruit of discussions with the GeoPandas team. GeoArrow is a packed columnar data format for the six fundamental geometry types: Point, MultiPoint, Lines, MultiLines, Polygons, and MultiPolygons. MultiGeometry is a possibility that may be implemented in the future. GeoArrow uses packed coordinate and offset columns to define objects, which enables very-fast copy between CPU, GPU, and NIC.
+
+Any data source that is loaded into cuSpatial via :func:`cuspatial.from_geopandas` can then take advantage of `cudf`'s GPU-accelerated Arrow I/O routines.
+
+Read more about GeoArrow format in :func:`GeoArrowBuffers<cuspatial.GeoArrowBuffers>`
+
+cuSpatial API Reference
+~~~~~~~~~~~~~~~~~~~~~~~
+
 .. toctree::
    :maxdepth: 2
    :caption: Contents:
 
    api.rst
-
+
+~~~~~~~~~~~~~~~~~~~~~~~
 
 Indices and tables
 ==================

python/cuspatial/cuspatial/__init__.py

@@ -21,7 +21,11 @@
     trajectory_bounding_boxes,
     trajectory_distances_and_speeds,
 )
+from .geometry.geoseries import GeoSeries
+from .geometry.geodataframe import GeoDataFrame
+from .geometry.geoarrowbuffers import GeoArrowBuffers
 from .io.shapefile import read_polygon_shapefile
+from .io.geopandas import from_geopandas
 
 __version__ = get_versions()["version"]
 del get_versions