Add built-in dictionary for dtype to default fill value

Author: gcaria

xarray/coding/variables.py

@@ -3,14 +3,27 @@
 from functools import partial
 from typing import Any, Hashable
 
-import netCDF4
 import numpy as np
 import pandas as pd
 
 from ..core import dtypes, duck_array_ops, indexing
 from ..core.pycompat import is_duck_dask_array
 from ..core.variable import Variable
 
+NP_TO_NC_TYPE = {
+    "S1": "\x00",
+    "i1": -127,
+    "u1": 255,
+    "i2": -32767,
+    "u2": 65535,
+    "i4": -2147483647,
+    "u4": 4294967295,
+    "i8": -9223372036854775806,
+    "u8": 18446744073709551614,
+    "f4": 9.969209968386869e36,
+    "f8": 9.969209968386869e36,
+}
+
 
 class SerializationWarning(RuntimeWarning):
     """Warnings about encoding/decoding issues in serialization."""
@@ -184,8 +197,14 @@ def decode(self, variable, name=None):
             pop_to(attrs, encoding, attr, name=name)
             for attr in ("missing_value", "_FillValue")
         ]
+        try:
+            default_fillvalue = NP_TO_NC_TYPE[np.dtype(variable.dtype).str[1:]]
+            raw_fill_values.append(default_fillvalue)
+        except KeyError:
+            warnings.warn(
+                f"A default fill value for dtype {variable.dtype} could not be found"
+            )
 
-        raw_fill_values.append(netCDF4.default_fillvals["f8"])
         if raw_fill_values:
             encoded_fill_values = {
                 fv
@@ -212,7 +231,6 @@ def decode(self, variable, name=None):
                     dtype=dtype,
                 )
                 data = lazy_elemwise_func(data, transform, dtype)
-
         return Variable(dims, data, attrs, encoding)
 
 

xarray/tests/test_conventions.py

@@ -1,7 +1,6 @@
 import contextlib
 import warnings
 
-import netCDF4
 import numpy as np
 import pandas as pd
 import pytest
@@ -234,7 +233,8 @@ def test_decode_cf_with_multiple_missing_values(self):
             assert "has multiple fill" in str(w[0].message)
 
     def test_decode_standard_missing_value(self):
-        original = Variable(["t"], [netCDF4.default_fillvals["f8"], 1, 2], {})
+
+        original = Variable(["t"], [coding.variables.NP_TO_NC_TYPE["f8"], 1, 2], {})
         expected = Variable(["t"], [np.nan, 1, 2], {})
         actual = conventions.decode_cf_variable("t", original)
         assert_identical(expected, actual)