Raise error for invalid reference date for encoding time units

doc/whats-new.rst

@@ -27,6 +27,8 @@ New Features
 - :py:meth:`Dataset.interp` now allows interpolation with non-numerical datatypes,
   such as booleans, instead of dropping them. (:issue:`4761` :pull:`5008`).
   By `Jimmy Westling <https://github.com/illviljan>`_.
+- Raise more informative error when decoding time variables with invalid reference dates.
+  (:issue:`5199`, :pull:`5288`). By `Giacomo Caria <https://github.com/gcaria>`_.
 
 Breaking changes
 ~~~~~~~~~~~~~~~~

xarray/coding/times.py

@@ -104,6 +104,8 @@ def _ensure_padded_year(ref_date):
     # No four-digit strings, assume the first digits are the year and pad
     # appropriately
     matches_start_digits = re.match(r"(\d+)(.*)", ref_date)
+    if not matches_start_digits:
+        raise ValueError(f"invalid reference date for time units: {ref_date}")
     ref_year, everything_else = [s for s in matches_start_digits.groups()]
     ref_date_padded = "{:04d}{}".format(int(ref_year), everything_else)
 

xarray/tests/test_coding_times.py

@@ -823,6 +823,12 @@ def test_encode_cf_datetime_overflow(shape):
     roundtrip = decode_cf_datetime(num, units, calendar)
     np.testing.assert_array_equal(dates, roundtrip)
 
+    with pytest.raises(ValueError, match="invalid time units"):
+        encode_cf_datetime(dates, units="days after 2000-01-01")
+
+    with pytest.raises(ValueError, match="invalid reference date"):
+        encode_cf_datetime(dates, units="days since NO_YEAR")
+
 
 def test_encode_cf_datetime_pandas_min():
     # GH 2623