Updates to conform to Geometry interface API (#407)

.travis.yml

@@ -6,10 +6,10 @@ os:
 env:
   global:
     - PYENV_VERSION=3.7
-    - PYTHON_VERSION=3.6
-    - PKG_TEST_PYTHON="--test-python=py36 --test-python=py27"
-    - CHANS_DEV="-c pyviz/label/dev -c bokeh/label/dev"
-    - CHANS_REL="-c pyviz -c bokeh"
+    - PYTHON_VERSION=3.7
+    - PKG_TEST_PYTHON="--test-python=py37 --test-python=py27"
+    - CHANS_DEV="-c pyviz/label/dev"
+    - CHANS_REL="-c pyviz"
     - LABELS_DEV="--label dev"
     - LABELS_REL="--label dev --label main"
     - MPLBACKEND="Agg"

geoviews/data/geom_dict.py

@@ -3,6 +3,8 @@
 
 import numpy as np
 from holoviews.core.data import Interface, DictInterface, MultiInterface
+from holoviews.core.data.interface import DataError
+from holoviews.core.data.spatialpandas import to_geom_dict
 from holoviews.core.dimension import OrderedDict as cyODict, dimension_name
 from holoviews.core.util import isscalar
 
@@ -32,6 +34,24 @@ def init(cls, eltype, data, kdims, vdims):
         dimensions = [dimension_name(d) for d in kdims + vdims]
         if isinstance(data, geom_types):
             data = {'geometry': data}
+        elif not isinstance(data, dict) or 'geometry' not in data:
+            xdim, ydim = kdims[:2]
+            from shapely.geometry import (
+                Point, LineString, Polygon, MultiPoint, MultiPolygon,
+                MultiLineString, LinearRing
+            )
+            data = to_geom_dict(eltype, data, kdims, vdims, GeomDictInterface)
+            geom = data.get('geom_type') or MultiInterface.geom_type(eltype)
+            poly = 'holes' in data or geom == 'Polygon'
+            if poly:
+                single_type, multi_type = Polygon, MultiPolygon
+            elif geom == 'Line':
+                single_type, multi_type = LineString, MultiLineString
+            elif geom == 'Ring':
+                single_type, multi_type = LinearRing, MultiPolygon
+            else:
+                single_type, multi_type = Point, MultiPoint
+            data['geometry'] = geom_from_dict(data, xdim.name, ydim.name, single_type, multi_type)
 
         if not cls.applies(data):
             raise ValueError("GeomDictInterface only handles dictionary types "
@@ -71,15 +91,43 @@ def init(cls, eltype, data, kdims, vdims):
 
     @classmethod
     def validate(cls, dataset, validate_vdims):
-        assert len([d for d in dataset.kdims + dataset.vdims
-                    if d.name not in dataset.data]) == 2
+        from shapely.geometry.base import BaseGeometry
+        geom_dims = cls.geom_dims(dataset)
+        if len(geom_dims) != 2:
+            raise DataError('Expected %s instance to declare two key '
+                            'dimensions corresponding to the geometry '
+                            'coordinates but %d dimensions were found '
+                            'which did not refer to any columns.'
+                            % (type(dataset).__name__, len(geom_dims)), cls)
+        elif 'geometry' not in dataset.data:
+            raise DataError("Could not find a 'geometry' column in the data.")
+        elif not isinstance(dataset.data['geometry'], BaseGeometry):
+            raise DataError("The 'geometry' column should be a shapely"
+                            "geometry type, found %s type instead." %
+                            type(dataset.data['geometry']).__name__)
 
     @classmethod
-    def dtype(cls, dataset, dimension):
-        name = dataset.get_dimension(dimension, strict=True).name
-        if name not in dataset.data:
-            return np.dtype('float') # Geometry dimension
-        return super(GeomDictInterface, cls).dtype(dataset, dimension)
+    def shape(cls, dataset):
+        return (cls.length(dataset), len(dataset.dimensions()))
+
+    @classmethod
+    def geom_type(cls, dataset):
+        from shapely.geometry import (
+            Polygon, MultiPolygon, LineString, MultiLineString, LinearRing
+        )
+        geom = dataset.data['geometry']
+        if isinstance(geom, (Polygon, MultiPolygon)):
+            return 'Polygon'
+        elif isinstance(geom, LinearRing):
+            return 'Ring'
+        elif isinstance(geom, (LineString, MultiLineString)):
+            return 'Line'
+        else:
+            return 'Point'
+
+    @classmethod
+    def geo_column(cls, dataset):
+        return 'geometry'
 
     @classmethod
     def has_holes(cls, dataset):
@@ -111,6 +159,13 @@ def dimension_type(cls, dataset, dim):
         values = dataset.data[name]
         return type(values) if isscalar(values) else values.dtype.type
 
+    @classmethod
+    def dtype(cls, dataset, dimension):
+        name = dataset.get_dimension(dimension, strict=True).name
+        if name in cls.geom_dims(dataset):
+            return np.dtype('float')
+        return Interface.dtype(dataset, dimension)
+
     @classmethod
     def range(cls, dataset, dim):
         dim = dataset.get_dimension(dim)
@@ -150,11 +205,82 @@ def values(cls, dataset, dim, expanded=True, flat=True, compute=True, keep_index
 
     @classmethod
     def select(cls, dataset, selection_mask=None, **selection):
-        raise NotImplementedError('select operation not implemented on geometries')
+        if cls.geom_dims(dataset):
+            data = cls.shape_mask(dataset, selection)
+        else:
+            data = dataset.data
+        if selection_mask is None:
+            selection_mask = cls.select_mask(dataset, selection)
+        empty = not selection_mask.sum()
+        dimensions = dataset.dimensions()
+        if empty:
+            return {d.name: np.array([], dtype=cls.dtype(dataset, d))
+                    for d in dimensions}
+        indexed = cls.indexed(dataset, selection)
+        new_data = {}
+        for k, v in data.items():
+            if k not in dimensions or isscalar(v):
+                new_data[k] = v
+            else:
+                new_data[k] = v[selection_mask]
+        if indexed and len(list(new_data.values())[0]) == 1 and len(dataset.vdims) == 1:
+            value = new_data[dataset.vdims[0].name]
+            return value if isscalar(value) else value[0]
+        return new_data
+
+    @classmethod
+    def shape_mask(cls, dataset, selection):
+        xdim, ydim = cls.geom_dims(dataset)
+        xsel = selection.pop(xdim.name, None)
+        ysel = selection.pop(ydim.name, None)
+        if xsel is None and ysel is None:
+            return dataset.data
+
+        from shapely.geometry import box
+
+        if xsel is None:
+            x0, x1 = cls.range(dataset, xdim)
+        elif isinstance(xsel, slice):
+            x0, x1 = xsel.start, xsel.stop
+        elif isinstance(xsel, tuple):
+            x0, x1 = xsel
+        else:
+            raise ValueError("Only slicing is supported on geometries, %s "
+                             "selection is of type %s."
+                             % (xdim, type(xsel).__name__))
+
+        if ysel is None:
+            y0, y1 = cls.range(dataset, ydim)
+        elif isinstance(ysel, slice):
+            y0, y1 = ysel.start, ysel.stop
+        elif isinstance(ysel, tuple):
+            y0, y1 = ysel
+        else:
+            raise ValueError("Only slicing is supported on geometries, %s "
+                             "selection is of type %s."
+                             % (ydim, type(ysel).__name__))
+
+        bounds = box(x0, y0, x1, y1)
+        geom = dataset.data['geometry']
+        geom = geom.intersection(bounds)
+        new_data = dict(dataset.data, geometry=geom)
+        return new_data
 
     @classmethod
     def iloc(cls, dataset, index):
-        raise NotImplementedError('iloc operation not implemented for geometries.')
+        from shapely.geometry import MultiPoint
+        rows, cols = index
+
+        data = dict(dataset.data)
+        geom = data['geometry']
+
+        if isinstance(geom, MultiPoint):
+            if isscalar(rows) or isinstance(rows, slice):
+                geom = geom[rows]
+            elif isinstance(rows, (set, list)):
+                geom = MultiPoint([geom[r] for r in rows])
+        data['geometry'] = geom
+        return data
 
     @classmethod
     def sample(cls, dataset, samples=[]):
@@ -169,5 +295,56 @@ def concat(cls, datasets, dimensions, vdims):
         raise NotImplementedError('concat operation not implemented for geometries.')
 
 
+def geom_from_dict(geom, xdim, ydim, single_type, multi_type):
+    from shapely.geometry import (
+        Point, LineString, Polygon, MultiPoint, MultiPolygon, MultiLineString
+    )
+    if (xdim, ydim) in geom:
+        xs, ys = np.asarray(geom.pop((xdim, ydim))).T
+    elif xdim in geom and ydim in geom:
+        xs, ys = geom.pop(xdim), geom.pop(ydim)
+    else:
+        raise ValueError('Could not find geometry dimensions')
+
+    xscalar, yscalar = isscalar(xs), isscalar(ys)
+    if xscalar and yscalar:
+        xs, ys = np.array([xs]), np.array([ys])
+    elif xscalar:
+        xs = np.full_like(ys, xs)
+    elif yscalar:
+        ys = np.full_like(xs, ys)
+    geom_array = np.column_stack([xs, ys])
+    splits = np.where(np.isnan(geom_array[:, :2].astype('float')).sum(axis=1))[0]
+    if len(splits):
+        split_geoms = [g[:-1] if i == (len(splits)-1) else g
+                       for i, g in enumerate(np.split(geom_array, splits+1))]
+    else:
+        split_geoms = [geom_array]
+    split_holes = geom.pop('holes', None)
+    if split_holes is not None and len(split_holes) != len(split_geoms):
+        raise DataError('Polygons with holes containing multi-geometries '
+                        'must declare a list of holes for each geometry.')
+
+    if single_type is Point:
+        if len(splits) > 1 or any(len(g) > 1 for g in split_geoms):
+            geom = MultiPoint(np.concatenate(split_geoms))
+        else:
+            geom = Point(*split_geoms[0])
+    elif len(splits):
+        if multi_type is MultiPolygon:
+            if split_holes is None:
+                split_holes = [[]]*len(split_geoms)
+            geom = MultiPolygon(list(zip(split_geoms, split_holes)))
+        else:
+            geom = MultiLineString(split_geoms)
+    elif single_type is Polygon:
+        if split_holes is None or not len(split_holes):
+            split_holes = [None]
+        geom = Polygon(split_geoms[0], split_holes[0])
+    else:
+        geom = LineString(split_geoms[0])
+    return geom
+
+
 MultiInterface.subtypes.insert(0, 'geom_dictionary')
 Interface.register(GeomDictInterface)