BUG: UndefinedUnitError: 'Celsius' is not defined in the unit registry

[pagecp]

• icclim version: 5.1.1-dev0
• Python version: 3.9.7

Description

Using E-OBS to calculate SU, I have this error:
UndefinedUnitError: 'Celsius' is not defined in the unit registry

Minimal reproducible example

summer_days = icclim.index(in_files=listOfFileNames, indice_name="SU", slice_mode='JJA')

Output received

2022-04-26 12:01:00,690 --- icclim 5.1.1-dev
2022-04-26 12:01:00,691 --- BEGIN EXECUTION
2022-04-26 12:01:00,691 Processing: 0%
2022-04-26 12:01:00,692 DEPRECATION_WARNING: `indice_name` is deprecated. Use `index_name` instead.
2022-04-26 12:01:00,766 Calculating climate index: SU
/data/home/globc/page/data/envs/icclimv5/lib/python3.9/site-packages/xclim/core/cfchecks.py:39: UserWarning: Variable has a non-conforming cell_methods: Got `time: mean`, which do not include the expected `time: maximum`
  _check_cell_methods(
---------------------------------------------------------------------------
UndefinedUnitError                        Traceback (most recent call last)
File ~/data/envs/icclimv5/lib/python3.9/site-packages/xclim/core/units.py:161, in units2pint(value)
    160 try:  # Pint compatible
--> 161     return units.parse_units(unit)
    162 except (
    163     pint.UndefinedUnitError,
    164     pint.DimensionalityError,
    165     AttributeError,
    166     TypeError,
    167 ):  # Convert from CF-units to pint-compatible

File ~/data/envs/icclimv5/lib/python3.9/site-packages/pint/registry.py:1194, in BaseRegistry.parse_units(self, input_string, as_delta, case_sensitive)
   1193     input_string = p(input_string)
-> 1194 units = self._parse_units(input_string, as_delta, case_sensitive)
   1195 return self.Unit(units)

File ~/data/envs/icclimv5/lib/python3.9/site-packages/pint/registry.py:1431, in NonMultiplicativeRegistry._parse_units(self, input_string, as_delta, case_sensitive)
   1429     as_delta = self.default_as_delta
-> 1431 return super()._parse_units(input_string, as_delta, case_sensitive)

File ~/data/envs/icclimv5/lib/python3.9/site-packages/pint/registry.py:1227, in BaseRegistry._parse_units(self, input_string, as_delta, case_sensitive)
   1226 for name in units:
-> 1227     cname = self.get_name(name, case_sensitive=case_sensitive)
   1228     value = units[name]

File ~/data/envs/icclimv5/lib/python3.9/site-packages/pint/registry.py:714, in BaseRegistry.get_name(self, name_or_alias, case_sensitive)
    713 if not candidates:
--> 714     raise UndefinedUnitError(name_or_alias)
    715 elif len(candidates) == 1:

UndefinedUnitError: 'Celsius' is not defined in the unit registry

During handling of the above exception, another exception occurred:

UndefinedUnitError                        Traceback (most recent call last)
Input In [16], in <cell line: 5>()
      3 summer_days = xr.Dataset
      4 #su = icclim.su(listOfFileNames, slice_mode='JJA')
----> 5 summer_days = icclim.index(in_files=listOfFileNames, indice_name="SU", slice_mode='JJA')

File ~/data/envs/icclimv5/lib/python3.9/site-packages/icclim/main.py:262, in index(in_files, index_name, var_name, slice_mode, time_range, out_file, threshold, callback, callback_percentage_start_value, callback_percentage_total, base_period_time_range, window_width, only_leap_years, ignore_Feb29th, interpolation, out_unit, netcdf_version, user_index, save_percentile, logs_verbosity, indice_name, user_indice, transfer_limit_Mbytes)
    260     result_ds = _compute_user_index_dataset(config=config, user_index=user_index)
    261 else:
--> 262     result_ds = _compute_ecad_index_dataset(
    263         config=config,
    264         index=index,
    265         threshold=threshold,
    266         current_history=ds.attrs.get("history", None),
    267     )
    268 if reset_coords_dict:
    269     result_ds = result_ds.rename(reset_coords_dict)

File ~/data/envs/icclimv5/lib/python3.9/site-packages/icclim/main.py:339, in _compute_ecad_index_dataset(config, index, threshold, current_history)
    337 else:
    338     config.threshold = threshold
--> 339     result_ds = _compute_ecad_index(index, config, current_history)
    340 return result_ds

File ~/data/envs/icclimv5/lib/python3.9/site-packages/icclim/main.py:391, in _compute_ecad_index(index, config, former_history)
    389 logging.info(f"Calculating climate index: {index.short_name}")
    390 result_ds = Dataset()
--> 391 res = index.compute(config)
    392 if isinstance(res, tuple):
    393     da, per = res

File ~/data/envs/icclimv5/lib/python3.9/site-packages/icclim/models/ecad_indices.py:140, in EcadIndex.<lambda>(c)
    130 VDTR = (
    131     "vDTR",
    132     lambda c: vdtr(c),
    133     IndexGroup.TEMPERATURE,
    134     [TAS_MAX, TAS_MIN],
    135 )
    137 # Heat
    138 SU = (
    139     "SU",
--> 140     lambda c: su(c),
    141     IndexGroup.HEAT,
    142     [TAS_MAX],
    143     [MODIFIABLE_THRESHOLD],
    144 )
    145 TR = (
    146     "TR",
    147     lambda c: tr(c),
   (...)
    150     [MODIFIABLE_THRESHOLD],
    151 )
    152 WSDI = (
    153     "WSDI",
    154     lambda c: wsdi(c),
   (...)
    157     [QUANTILE_BASED, MODIFIABLE_QUANTILE_WINDOW, MODIFIABLE_THRESHOLD],
    158 )

File ~/data/envs/icclimv5/lib/python3.9/site-packages/icclim/ecad_functions.py:145, in su(config)
    139     result = atmos.maximum_consecutive_dry_days(
    140         config.pr.study_da, thresh="1.0 mm/day", freq=config.freq.panda_freq
    141     )
    142     return result
--> 145 def su(config: IndexConfig) -> DataArray:
    146     return _compute_threshold_index(
    147         da=config.tasmax.study_da,
    148         threshold=25.0 if config.threshold is None else config.threshold,
    149         freq=config.freq,
    150         xclim_index_fun=atmos.tx_days_above,
    151     )
    154 def tr(config: IndexConfig) -> DataArray:

File ~/data/envs/icclimv5/lib/python3.9/site-packages/icclim/ecad_functions.py:878, in _compute_threshold_index(da, threshold, freq, xclim_index_fun)
    871     result = xclim_index_fun(da, thresh=f"{threshold} °C", freq=freq.panda_freq)
    872     return result
    875 def _compute_spell_duration(
    876     cf_var: CfVariable,
    877     freq: str,
--> 878     per_window: int,
    879     per_thresh: float,
    880     per_interpolation: QuantileInterpolation,
    881     min_spell_duration: int,
    882     save_percentile: bool,
    883     callback: Callable,
    884     xclim_index_fun: Callable,
    885 ) -> Tuple[DataArray, Optional[DataArray]]:
    886     per = _compute_percentile_doy(
    887         cf_var.reference_da,
    888         per_thresh,
   (...)
    891         callback,
    892     )
    893     run_bootstrap = _can_run_bootstrap(cf_var)

File ~/data/envs/icclimv5/lib/python3.9/site-packages/xclim/core/indicator.py:790, in Indicator.__call__(self, *args, **kwds)
    788     elif nm not in compute_das and nm in params:
    789         kwargs[nm] = params[nm]
--> 790 outs = self.compute(**compute_das, **kwargs, **var_kwargs)
    792 if isinstance(outs, DataArray):
    793     outs = [outs]

File <boltons.funcutils.FunctionBuilder-36>:2, in tx_days_above(tasmax, thresh, freq)

File ~/data/envs/icclimv5/lib/python3.9/site-packages/xclim/core/units.py:688, in declare_units.<locals>.dec.<locals>.wrapper(*args, **kwargs)
    686 bound_args = sig.bind(*args, **kwargs)
    687 for name, val in bound_args.arguments.items():
--> 688     check_units(val, bound_units.arguments.get(name, None))
    690 out = func(*args, **kwargs)
    692 # Perform very basic sanity check on the output.
    693 # Indice are responsible for unit management.
    694 # If this fails, it's a developer's error.

File <boltons.funcutils.FunctionBuilder-1>:2, in check_units(val, dim)

File ~/data/envs/icclimv5/lib/python3.9/site-packages/xclim/core/options.py:116, in datacheck.<locals>.run_check(*args, **kwargs)
    114 @wraps(func)
    115 def run_check(*args, **kwargs):
--> 116     return _run_check(func, DATA_VALIDATION, *args, **kwargs)

File ~/data/envs/icclimv5/lib/python3.9/site-packages/xclim/core/options.py:106, in _run_check(func, option, *args, **kwargs)
    104 """Run function and customize exception handling based on option."""
    105 try:
--> 106     func(*args, **kwargs)
    107 except ValidationError as err:
    108     raise_warn_or_log(err, OPTIONS[option], stacklevel=4)

File ~/data/envs/icclimv5/lib/python3.9/site-packages/xclim/core/units.py:634, in check_units(val, dim)
    632     val_units = str2pint(val)
    633 else:  # a DataArray
--> 634     val_units = units2pint(val)
    635 val_dim = val_units.dimensionality
    637 if val_dim == expected:

File ~/data/envs/icclimv5/lib/python3.9/site-packages/xclim/core/units.py:168, in units2pint(value)
    161     return units.parse_units(unit)
    162 except (
    163     pint.UndefinedUnitError,
    164     pint.DimensionalityError,
    165     AttributeError,
    166     TypeError,
    167 ):  # Convert from CF-units to pint-compatible
--> 168     return units.parse_units(_transform(unit))

File ~/data/envs/icclimv5/lib/python3.9/site-packages/pint/registry.py:1194, in BaseRegistry.parse_units(self, input_string, as_delta, case_sensitive)
   1192 for p in self.preprocessors:
   1193     input_string = p(input_string)
-> 1194 units = self._parse_units(input_string, as_delta, case_sensitive)
   1195 return self.Unit(units)

File ~/data/envs/icclimv5/lib/python3.9/site-packages/pint/registry.py:1431, in NonMultiplicativeRegistry._parse_units(self, input_string, as_delta, case_sensitive)
   1428 if as_delta is None:
   1429     as_delta = self.default_as_delta
-> 1431 return super()._parse_units(input_string, as_delta, case_sensitive)

File ~/data/envs/icclimv5/lib/python3.9/site-packages/pint/registry.py:1227, in BaseRegistry._parse_units(self, input_string, as_delta, case_sensitive)
   1225 many = len(units) > 1
   1226 for name in units:
-> 1227     cname = self.get_name(name, case_sensitive=case_sensitive)
   1228     value = units[name]
   1229     if not cname:

File ~/data/envs/icclimv5/lib/python3.9/site-packages/pint/registry.py:714, in BaseRegistry.get_name(self, name_or_alias, case_sensitive)
    712 candidates = self.parse_unit_name(name_or_alias, case_sensitive)
    713 if not candidates:
--> 714     raise UndefinedUnitError(name_or_alias)
    715 elif len(candidates) == 1:
    716     prefix, unit_name, _ = candidates[0]

UndefinedUnitError: 'Celsius' is not defined in the unit registry

[bzah]

MetPy seemed to had a similar issue and raised it to Pint (the library handling unit): hgrecco/pint#1081

It seems "Celsius" should be written in lowercase "celsius" according to standards.
However, xclim already offer some additional aliases such as "C" or "deg_C". We could add either on icclim on xclim some additional aliases.

There is also some effort to backport to the cf_xarray units handling but I'm not sure how is it going: xarray-contrib/cf-xarray#284
We could try to contribute there as well.

As a workaround you could rewrite the "units" in your dataset:

ds = xarray.open_mfdataset(listOfFileNames) # you might need to glob.glob it before, I don't remember
ds.attrs["units"] = "celsius"
ds.to_netcdf("obs_fixed.nc")

[bzah]

I think you can also dynamically update the registry with:

from xclim.core.units import units
units.define("@alias degC = C = deg_C = Celsius")

# ...

icclim.indice(blablabla)
``

[pagecp]

Just quick note, is it spelled Celsius (not Celcius) :)

[bzah]

oopsi, fixing it.

[bzah]

I'm closing this as 2 PR are in progress in xclim and cf-xarray:

• ENH: Add "Celsius" alias xarray-contrib/cf-xarray#336
• Enh: Add Celsius Ouranosinc/xclim#1068
  and we have a workaround for icclim users for now.