ENH: setncattr_string support (GH2044)

Author: dnowacki-usgs

doc/whats-new.rst

@@ -34,6 +34,10 @@ v0.10.4 (unreleased)
 Enhancements
 ~~~~~~~~~~~~
 
+- Support reading and writing lists of strings as attributes to netCDF
+  (:issue:`2044`).
+  By `Dan Nowacki <https://github.com/dnowacki-usgs>`_.
+
 Bug fixes
 ~~~~~~~~~
 

xarray/backends/netCDF4_.py

@@ -229,6 +229,31 @@ def _disable_auto_decode_group(ds):
         _disable_auto_decode_variable(var)
 
 
+def _is_list_of_strings(value):
+    if (np.asarray(value).dtype.kind in ['U', 'S'] and
+            np.asarray(value).size > 1):
+        return True
+    else:
+        return False
+
+
+def _set_nc_attribute(obj, key, value):
+    if _is_list_of_strings(value):
+        # encode as NC_STRING if attr is list of strings
+        try:
+            obj.setncattr_string(key, value)
+        except AttributeError:
+            # Inform users with old netCDF that does not support
+            # NC_STRING that we can't serialize lists of strings
+            # as attrs
+            msg = ('Attributes which are lists of strings are not '
+                   'supported with this version of netCDF. Please '
+                   'upgrade to netCDF4-python 1.2.4 or greater.')
+            raise AttributeError(msg)
+    else:
+        obj.setncattr(key, value)
+
+
 class NetCDF4DataStore(WritableCFDataStore, DataStorePickleMixin):
     """Store for reading and writing data via the Python-NetCDF4 library.
 
@@ -347,7 +372,7 @@ def set_attribute(self, key, value):
         with self.ensure_open(autoclose=False):
             if self.format != 'NETCDF4':
                 value = encode_nc3_attr_value(value)
-            self.ds.setncattr(key, value)
+            _set_nc_attribute(self.ds, key, value)
 
     def set_variables(self, *args, **kwargs):
         with self.ensure_open(autoclose=False):
@@ -402,7 +427,7 @@ def prepare_variable(self, name, variable, check_encoding=False,
         for k, v in iteritems(attrs):
             # set attributes one-by-one since netCDF4<1.0.10 can't handle
             # OrderedDict as the input to setncatts
-            nc4_var.setncattr(k, v)
+            _set_nc_attribute(nc4_var, k, v)
 
         target = NetCDF4ArrayWrapper(name, self)
 

xarray/tests/test_backends.py

@@ -1144,6 +1144,23 @@ def test_88_character_filename_segmentation_fault(self):
                     # Need to construct 88 character filepath
                     xr.Dataset().to_netcdf('a' * (88 - len(os.getcwd()) - 1))
 
+    def test_setncattr_string(self):
+        list_of_strings = ['list', 'of', 'strings']
+        one_element_list_of_strings = ['one element']
+        one_string = 'one string'
+        attrs = {'foo': list_of_strings,
+                 'bar': one_element_list_of_strings,
+                 'baz': one_string}
+        ds = Dataset({'x': ('y', [1, 2, 3], attrs)},
+                     attrs=attrs)
+
+        with self.roundtrip(ds) as actual:
+            for totest in [actual, actual['x']]:
+                assert_array_equal(list_of_strings, totest.attrs['foo'])
+                assert_array_equal(one_element_list_of_strings,
+                                   totest.attrs['bar'])
+                assert one_string == totest.attrs['baz']
+
 
 class NetCDF4DataStoreAutocloseTrue(NetCDF4DataTest):
     autoclose = True