Resolve #1822

CHANGES.txt

@@ -1,9 +1,12 @@
 Changes
 =======
 
-1.1.1 (2019-11-12)
+1.1.1 (2019-11-13)
 ------------------
 
+- Calling a dataset's sample method with coordinates outside the extent of a
+  dataset with a nodata value of None has raised a TypeError (#1822). Now, it
+  gives the values we would get from a boundless read of the dataset's values.
 - Use new set_proj_search_path() function to set the PROJ data search path. For
   GDAL versions before 3.0 this sets the PROJ_LIB environment variable. For
   GDAL version 3.0 this calls OSRSetPROJSearchPaths(), which overrides

rasterio/_io.pyx

@@ -762,30 +762,35 @@ cdef class DatasetReaderBase(DatasetBase):
                 mask = mask | self.read_masks(i, **kwargs)
             return mask
 
-    def sample(self, xy, indexes=None):
+    def sample(self, xy, indexes=None, masked=False):
         """Get the values of a dataset at certain positions
 
         Values are from the nearest pixel. They are not interpolated.
 
         Parameters
         ----------
-        xy : iterable, pairs of floats
-            A sequence or generator of (x, y) pairs.
-
-        indexes : list of ints or a single int, optional
-            If `indexes` is a list, the result is a 3D array, but is
-            a 2D array if it is a band index number.
+        xy : iterable
+            Pairs of x, y coordinates (floats) in the dataset's
+            reference system.
+        indexes : int or list of int
+            Indexes of dataset bands to sample.
+        masked : bool, default: False
+            Whether to mask samples that fall outside the extent of the
+            dataset.
 
         Returns
-        -------
-        Iterable, yielding dataset values for the specified `indexes`
-        as an ndarray.
+        ------
+        iterable
+            Arrays of length equal to the number of specified indexes
+            containing the dataset values for the bands corresponding to
+            those indexes.
+
         """
         # In https://github.com/mapbox/rasterio/issues/378 a user has
         # found what looks to be a Cython generator bug. Until that can
         # be confirmed and fixed, the workaround is a pure Python
         # generator implemented in sample.py.
-        return sample_gen(self, xy, indexes)
+        return sample_gen(self, xy, indexes=indexes, masked=masked)
 
 
 @contextmanager

rasterio/sample.py

@@ -5,8 +5,29 @@ import numpy
 from rasterio.windows import Window
 
 
-def sample_gen(dataset, xy, indexes=None):
-    """Generator for sampled pixels"""
+def sample_gen(dataset, xy, indexes=None, masked=False):
+    """Sample pixels from a dataset
+
+    Parameters
+    ----------
+    dataset : rasterio Dataset
+        Opened in "r" mode.
+    xy : iterable
+        Pairs of x, y coordinates in the dataset's reference system.
+    indexes : int or list of int
+        Indexes of dataset bands to sample.
+    masked : bool, default: False
+        Whether to mask samples that fall outside the extent of the
+        dataset.
+
+    Yields
+    ------
+    array
+        A array of length equal to the number of specified indexes
+        containing the dataset values for the bands corresponding to
+        those indexes.
+
+    """
     index = dataset.index
     read = dataset.read
 
@@ -15,9 +36,9 @@ def sample_gen(dataset, xy, indexes=None):
 
     for x, y in xy:
         row_off, col_off = index(x, y)
-        if row_off < 0 or col_off < 0:
-            yield numpy.ones((dataset.count,), dtype=dataset.dtypes[0]) * dataset.nodata
-        else:
+#        if row_off < 0 or col_off < 0:
+#            yield numpy.ones((dataset.count,), dtype=dataset.dtypes[0]) * dataset.nodata
+#        else:
         window = Window(col_off, row_off, 1, 1)
-            data = read(indexes, window=window, masked=False)
+        data = read(indexes, window=window, masked=masked, boundless=True)
         yield data[:, 0, 0]

tests/test_sampling.py

@@ -1,3 +1,5 @@
+import numpy
+
 import rasterio
 
 
@@ -13,6 +15,24 @@ def test_sampling_beyond_bounds():
         assert list(data) == [0, 0, 0]
 
 
+def test_sampling_beyond_bounds_no_nodata():
+    with rasterio.open('tests/data/RGB2.byte.tif') as src:
+        data = next(src.sample([(-10, 2719200.0)]))
+        assert list(data) == [0, 0, 0]
+
+
+def test_sampling_beyond_bounds_masked():
+    with rasterio.open('tests/data/RGBA.byte.tif') as src:
+        data = next(src.sample([(-10, 2719200.0)], masked=True))
+        assert list(data.mask) == [True, True, True, False]
+
+
+def test_sampling_beyond_bounds_nan():
+    with rasterio.open('tests/data/float_nan.tif') as src:
+        data = next(src.sample([(-10, 0.0)]))
+        assert numpy.isnan(data)
+
+
 def test_sampling_indexes():
     with rasterio.open('tests/data/RGB.byte.tif') as src:
         data = next(src.sample([(220650.0, 2719200.0)], indexes=[2]))