Spatial module replaced with LatLon fields.

Makefile

@@ -6,9 +6,8 @@ clean:
 	rm -rf dist lupyne.egg-info
 
 html:
-	make -C docs $@ SPHINXOPTS=-W SPHINXBUILD=sphinx-build
+	make -C docs $@ SPHINXOPTS=-W
 	rst2$@.py README.rst docs/_build/README.$@
-	python3 -m examples.spatial > docs/_build/spatial.kml
 
 dist: html
 	python3 setup.py sdist bdist_wheel

docs/engine.rst

@@ -4,9 +4,8 @@ engine
 
   * `TokenFilter`_, `Analyzer`_
   * `IndexSearcher`_, `MultiSearcher`_, `IndexWriter`_, `Indexer`_
-  * `Document`_, `Field`_, `NestedField`_, `NumericField`_, `DateTimeField`_
+  * `Document`_, `Field`_, `NestedField`_, `NumericField`_, `DateTimeField`_, `SpatialField`_
   * `Query`_
-  * `PointField`_
 
 
 analyzers
@@ -166,6 +165,12 @@ DateTimeField
   :show-inheritance:
   :members:
 
+SpatialField
+^^^^^^^^^^^^^
+.. autoclass:: SpatialField
+  :show-inheritance:
+  :members:
+
 
 queries
 ---------
@@ -204,26 +209,3 @@ SpellParser
   .. attribute:: searcher
 
     `IndexSearcher`_
-
-
-spatial
----------
-.. automodule:: lupyne.engine.spatial
-
-Point
-^^^^^^^^^^^^^
-.. autoclass:: Point
-  :show-inheritance:
-  :members:
-
-Tile
-^^^^^^^^^^^^^
-.. autoclass:: Tile
-  :show-inheritance:
-  :members:
-
-PointField
-^^^^^^^^^^^^^
-.. autoclass:: PointField
-  :show-inheritance:
-  :members:

docs/examples.rst

@@ -25,7 +25,3 @@ grouping
 server
 ---------
 .. literalinclude:: ../examples/server.py
-
-spatial
----------
-.. literalinclude:: ../examples/spatial.py

examples/searching.py

@@ -9,21 +9,20 @@
 
 The general solution is to index the term values into a prefix tree.
 Then each query can expand to only values of the appropriate granularity.
-Lucene's NumericFields encode numbers to be sortable, so it is also able to cluster prefixes into the same field.
+Lucene's Point fields encode numbers to be sortable, so it is also able to cluster prefixes into the same field.
 Whereas Lupyne's NestedField assumes the value is already a sortable string, so different fields must be used to cluster the prefixes.
 There are trade-offs to each approach:
- * NumericFields support range queries natively, but must translate prefix queries.
+ * Point fields support range queries natively, but must translate prefix queries.
  * NestedFields support prefix queries optimally, but must translate range queries.
- * NumericFields only support numbers, and result in unreadable values in the index.
+ * Point fields only support numbers, and result in unreadable values in the index.
  * NestedFields support any searchable values, but pollute the field namespace.
 
-Lupyne PointFields and DateTimeFields are implemented as NumericFields since both are easily encoded as numbers.
+Lupyne SpatialFields and DateTimeFields are implemented as lucene Point fields.
 NestedFields could still be used however, as demonstrated on dates below.
 """
 
 from datetime import date
 import lucene
-from org.apache.lucene import search
 from lupyne import engine
 assert lucene.getVMEnv() or lucene.initVM()
 
@@ -40,7 +39,7 @@
 indexer.set('incorporated', engine.DateTimeField)
 indexer.set('year-month-day', engine.NestedField, sep='-')
 indexer.set('population', engine.NumericField)
-indexer.set('point', engine.PointField, precision=10)
+indexer.set('point', engine.SpatialField)
 # assigned fields can have a different key from their underlying field name
 indexer.fields['location'] = engine.NestedField('state.city')
 
@@ -69,9 +68,8 @@
 assert [hit['city'] for hit in indexer.search(query)] == ['San Francisco', 'Portland']
 
 cities = ['San Francisco', 'Los Angeles', 'Portland']
-for index, distance in enumerate([1e3, 1e5, 2e5, 1e6]):
+for index, distance in enumerate([1e3, 1e5, 7e5, 1e6]):
     query = indexer.fields['point'].within(-122.4, 37.7, distance=distance)
-    assert isinstance(query, search.BooleanQuery) and len(list(query)) <= 4
     assert {hit['city'] for hit in indexer.search(query)} == set(cities[:index])
 
 query = indexer.fields['location'].prefix('CA.San')

lupyne/engine/__init__.py

@@ -8,9 +8,8 @@
 import lucene  # flake8: noqa
 from .analyzers import Analyzer, TokenFilter
 from .queries import Query
-from .documents import Document, Field, NestedField, NumericField, DateTimeField
+from .documents import Document, Field, NestedField, NumericField, DateTimeField, SpatialField
 from .indexers import IndexSearcher, MultiSearcher, IndexWriter, Indexer
-from .spatial import PointField
 
 version = tuple(map(int, lucene.VERSION.split('.')))
 assert version >= (6,), version

lupyne/engine/documents.py

@@ -238,6 +238,29 @@ def within(self, days=0, weeks=0, utc=True, **delta):
         return self.duration(date, days, weeks=weeks, **delta)
 
 
+class SpatialField(NumericField):
+    """Geospatial points, indexed with optional docvalues."""
+    def items(self, *points):
+        """Generate lucene LatLon fields from points (lng, lat)."""
+        for lng, lat in points:
+            yield document.LatLonPoint(self.name, lat, lng)
+        if self.docvalues:
+            for lng, lat in points:
+                yield document.LatLonDocValuesField(self.name, lat, lng)
+
+    def within(self, lng, lat, distance):
+        """Return range queries for any tiles which could be within distance of given point.
+
+        :param lng,lat: point
+        :param distance: search radius in meters
+        """
+        return document.LatLonPoint.newDistanceQuery(self.name, lat, lng, distance)
+
+    def distances(self, lng, lat):
+        """Return distance SortField."""
+        return document.LatLonDocValuesField.newDistanceSort(self.name, lat, lng)
+
+
 class Document(dict):
     """Multimapping of field names to values, but default getters return the first value."""
     def __init__(self, doc):

lupyne/engine/indexers.py

@@ -19,7 +19,6 @@
 from .analyzers import Analyzer
 from .queries import suppress, Query, DocValues, SpellParser
 from .documents import Field, Document, Hits, GroupingSearch
-from .spatial import Distances
 from ..utils import long, Atomic, SpellChecker
 
 for cls in (analysis.TokenStream, lucene.JArray_byte):
@@ -183,11 +182,6 @@ def docvalues(self, name, type=None):
         method = getattr(index.MultiDocValues, 'get{}Values'.format(docValuesType))
         return getattr(DocValues, docValuesType)(method(self.indexReader, name), len(self), type)
 
-    def distances(self, lng, lat, lngfield, latfield):
-        """Return distance calculator from given point and DocValue lng/lat fields."""
-        arrays = (self.docvalues(field, float) for field in (lngfield, latfield))
-        return Distances(lng, lat, *arrays)
-
     def copy(self, dest, query=None, exclude=None, merge=0):
         """Copy the index to the destination directory.