Use add_note() to annotate exceptions when encoding fails (#10464)

Co-authored-by: Deepak Cherian <dcherian@users.noreply.github.com>

Author: shoyer

doc/whats-new.rst

@@ -14,6 +14,8 @@ New Features
 ~~~~~~~~~~~~
 - Expose :py:class:`~xarray.indexes.RangeIndex`, and :py:class:`~xarray.indexes.CoordinateTransformIndex` as public api
   under the ``xarray.indexes`` namespace. By `Deepak Cherian <https://github.com/dcherian>`_.
+- Better error messages when encoding data to be written to disk fails (:pull:`10464`).
+  By `Stephan Hoyer <https://github.com/shoyer>`_
 
 Breaking changes
 ~~~~~~~~~~~~~~~~

xarray/backends/common.py

@@ -389,9 +389,23 @@ def encode(self, variables, attributes):
         attributes : dict-like
 
         """
-        variables = {k: self.encode_variable(v, name=k) for k, v in variables.items()}
-        attributes = {k: self.encode_attribute(v) for k, v in attributes.items()}
-        return variables, attributes
+        encoded_variables = {}
+        for k, v in variables.items():
+            try:
+                encoded_variables[k] = self.encode_variable(v)
+            except Exception as e:
+                e.add_note(f"Raised while encoding variable {k!r} with value {v!r}")
+                raise
+
+        encoded_attributes = {}
+        for k, v in attributes.items():
+            try:
+                encoded_attributes[k] = self.encode_attribute(v)
+            except Exception as e:
+                e.add_note(f"Raised while encoding attribute {k!r} with value {v!r}")
+                raise
+
+        return encoded_variables, encoded_attributes
 
     def encode_variable(self, v, name=None):
         """encode one variable"""
@@ -641,9 +655,7 @@ def encode(self, variables, attributes):
         variables = {
             k: ensure_dtype_not_object(v, name=k) for k, v in variables.items()
         }
-        variables = {k: self.encode_variable(v, name=k) for k, v in variables.items()}
-        attributes = {k: self.encode_attribute(v) for k, v in attributes.items()}
-        return variables, attributes
+        return super().encode(variables, attributes)
 
 
 class BackendEntrypoint:

xarray/conventions.py

@@ -792,7 +792,13 @@ def cf_encoder(variables: T_Variables, attributes: T_Attrs):
     # add encoding for time bounds variables if present.
     _update_bounds_encoding(variables)
 
-    new_vars = {k: encode_cf_variable(v, name=k) for k, v in variables.items()}
+    new_vars = {}
+    for k, v in variables.items():
+        try:
+            new_vars[k] = encode_cf_variable(v, name=k)
+        except Exception as e:
+            e.add_note(f"Raised while encoding variable {k!r} with value {v!r}")
+            raise
 
     # Remove attrs from bounds variables (issue #2921)
     for var in new_vars.values():

xarray/tests/test_backends_common.py

@@ -1,9 +1,13 @@
 from __future__ import annotations
 
+import re
+
 import numpy as np
 import pytest
 
+import xarray as xr
 from xarray.backends.common import _infer_dtype, robust_getitem
+from xarray.tests import requires_scipy
 
 
 class DummyFailure(Exception):
@@ -43,3 +47,16 @@ def test_robust_getitem() -> None:
 def test_infer_dtype_error_on_mixed_types(data):
     with pytest.raises(ValueError, match="unable to infer dtype on variable"):
         _infer_dtype(data, "test")
+
+
+@requires_scipy
+def test_encoding_failure_note():
+    # Create an arbitrary value that cannot be encoded in netCDF3
+    ds = xr.Dataset({"invalid": np.array([2**63 - 1], dtype=np.int64)})
+    with pytest.raises(
+        ValueError,
+        match=re.escape(
+            "Raised while encoding variable 'invalid' with value <xarray.Variable"
+        ),
+    ):
+        ds.to_netcdf()