Pytest fails depending on Grib Definitions

[MeraX]

What happened?

I noticed 4 tests in test_height_level_humidity.py were failing, depending on the used eccodes definition path.

================================================== short test summary info ==================================================
FAILED test_height_level_humidity.py::test_height_level_specific_humidity_to_relative_humidity_from_file - TypeError: '<' not supported between instances of 'str' and 'int'
FAILED test_height_level_humidity.py::test_relative_humidity_to_specific_humidity_from_file - TypeError: '<' not supported between instances of 'str' and 'int'
FAILED test_height_level_humidity.py::test_specific_humidity_to_dewpoint_from_file - TypeError: '<' not supported between instances of 'str' and 'int'
FAILED test_height_level_humidity.py::test_dewpoint_to_specific_humidity_from_file - TypeError: '<' not supported between instances of 'str' and 'int'
========================================== 4 failed, 8 passed, 1 warning in 4.23s ===========================================

I was wondering if it were possible to define these agnostic of the particular definition of shortnames.

What are the steps to reproduce the bug?

Run pytest using he following grib definitions: https://opendata.dwd.de/weather/lib/grib/eccodes_definitions.edzw-2.38.3-1.tar.bz2

Version

0.1.17

Platform (OS and architecture)

NA

Relevant log output

============================= test session starts ==============================
platform linux -- Python 3.11.11, pytest-8.4.2, pluggy-1.6.0 -- /shared/data/majacob/datasets/.venv/bin/python3
cachedir: .pytest_cache
rootdir: /shared/data/majacob/datasets/anemoi-transform
configfile: pyproject.toml
plugins: mock-3.15.1, xdist-3.8.0, skip-slow-0.0.5
collecting ... collected 1 item

test_height_level_humidity.py::test_dewpoint_to_specific_humidity_from_file FAILED [100%]

=================================== FAILURES ===================================
_________________ test_dewpoint_to_specific_humidity_from_file _________________

test_source = <function test_source.<locals>._source at 0x14c15fae6840>

    @skip_if_offline
    def test_dewpoint_to_specific_humidity_from_file(test_source):
        source = test_source("anemoi-transform/filters/input_single_level_specific_humidity_to_relative_humidity.grib")
        input_dewpoint_temperature = test_source(
            "anemoi-transform/filters/single_level_dewpoint_temperature.npy"
        ).ds.to_numpy()
        md = source.ds.sel(param="2d")[0].metadata()
        ds = source.ds.sel(param=["2sh", "2t", "sp", "q", "t"])
        ds += ekd.FieldList.from_array(input_dewpoint_temperature, md)
        source.ds = ds
    
        d_to_q_height = filter_registry.create(
            "d_to_q_height",
            height=2,
            specific_humidity_at_height_level="2q",
            dewpoint_temperature_at_height_level="2d",
            surface_pressure="sp",
            specific_humidity_at_model_levels="q",
            temperature_at_model_levels="t",
            model_level_AB=AB_coefficients_files,
        )
    
        pipeline = source | d_to_q_height
    
        input_fields = collect_fields_by_param(source)
>       output_fields = collect_fields_by_param(pipeline)
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

test_height_level_humidity.py:523: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
utils/__init__.py:20: in collect_fields_by_param
    for field in pipeline:
../src/anemoi/transform/workflow.py:28: in __iter__
    return iter(self(None))
                ^^^^^^^^^^
../src/anemoi/transform/workflow.py:43: in __call__
    return self.forward(data)
           ^^^^^^^^^^^^^^^^^^
../src/anemoi/transform/workflows/pipeline.py:47: in forward
    data = filter.forward(data)
           ^^^^^^^^^^^^^^^^^^^^
../src/anemoi/transform/transform.py:203: in forward
    return self.filter.backward(x)
           ^^^^^^^^^^^^^^^^^^^^^^^
../src/anemoi/transform/filters/matching.py:321: in backward
    return self._transform(
../src/anemoi/transform/filters/matching.py:356: in _transform
    for matching in grouping.iterate(data, other=result.append):
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

self = <anemoi.transform.grouping.GroupByParamVertical object at 0x14c14a225f90>
data = MultiFieldList(MaskIndex(GRIBReader(/hpc/gtmp/majacob/.jtmp.gpnl101.20251023.141832.126848/pytest-of-majacob/pytest-16...ity_to_relative_humidity.grib),[0, 1, 2, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19]),SimpleFieldList(1))

    def iterate(self, data: list[Any], *, other: Callable[[Any], None] = _lost) -> Iterator[tuple[Any, ...]]:
        """Iterate over the data and group fields by parameters.
    
        Parameters
        ----------
        data : list of Any
            List of data fields to group.
        other : callable, optional
            Function to call for fields that do not match the parameters, by default _lost.
    
        Returns
        -------
        Iterator[Tuple[Any, ...]]
            Iterator yielding tuples of grouped fields.
        """
        self._get_groups(data, other=other)
        for _, group in self.groups.items():
            if len(group) != len(self.params):
                for p in data:
                    print(p)
>               raise ValueError(f"Missing component. Want {sorted(self.params)}, got {sorted(self.groups.keys())}")
                                                                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^
E               TypeError: '<' not supported between instances of 'str' and 'int'

../src/anemoi/transform/grouping/__init__.py:115: TypeError
----------------------------- Captured stdout call -----------------------------
GribField(2sh,None,20220101,0,2,0)
GribField(sp,None,20220101,0,2,0)
GribField(2t,None,20220101,0,2,0)
GribField(t,130,20220101,0,2h,None)
GribField(q,130,20220101,0,2h,None)
GribField(t,131,20220101,0,2h,None)
GribField(q,131,20220101,0,2h,None)
GribField(t,132,20220101,0,2h,None)
GribField(q,132,20220101,0,2h,None)
GribField(t,133,20220101,0,2h,None)
GribField(q,133,20220101,0,2h,None)
GribField(t,134,20220101,0,2h,None)
GribField(q,134,20220101,0,2h,None)
GribField(t,135,20220101,0,2h,None)
GribField(q,135,20220101,0,2h,None)
GribField(t,136,20220101,0,2h,None)
GribField(q,136,20220101,0,2h,None)
GribField(t,137,20220101,0,2h,None)
GribField(q,137,20220101,0,2h,None)
ArrayField(2d,None,20220101,0,2,0)
----------------------------- Captured stderr call -----------------------------

input_single_level_specific_humidity_to_relative_humidity.grib:   0%|          | 0.00/63.4M [00:00<?, ?B/s]
input_single_level_specific_humidity_to_relative_humidity.grib:   2%|▏         | 1.00M/63.4M [00:00<00:18, 3.49MB/s]
input_single_level_specific_humidity_to_relative_humidity.grib:  17%|█▋        | 11.0M/63.4M [00:00<00:01, 35.9MB/s]
input_single_level_specific_humidity_to_relative_humidity.grib:  36%|███▋      | 23.0M/63.4M [00:00<00:00, 62.7MB/s]
input_single_level_specific_humidity_to_relative_humidity.grib:  55%|█████▌    | 35.0M/63.4M [00:00<00:00, 80.2MB/s]
input_single_level_specific_humidity_to_relative_humidity.grib:  74%|███████▍  | 47.0M/63.4M [00:00<00:00, 91.9MB/s]
input_single_level_specific_humidity_to_relative_humidity.grib:  93%|█████████▎| 59.0M/63.4M [00:00<00:00, 99.8MB/s]
                                                                                                                    

single_level_dewpoint_temperature.npy:   0%|          | 0.00/12.7M [00:00<?, ?B/s]
single_level_dewpoint_temperature.npy:   8%|▊         | 1.00M/12.7M [00:00<00:03, 4.04MB/s]
single_level_dewpoint_temperature.npy:  79%|███████▉  | 10.0M/12.7M [00:00<00:00, 35.7MB/s]
                                                                                           
------------------------------ Captured log call -------------------------------
WARNING  anemoi.transform.filters.matching:matching.py:211 Some inputs will not be returned because filter direction is backward: {'specific_humidity_at_height_level'}
=============================== warnings summary ===============================
../../.venv/lib64/python3.11/site-packages/_pytest/config/__init__.py:833
  /shared/data/majacob/datasets/.venv/lib64/python3.11/site-packages/_pytest/config/__init__.py:833: PytestAssertRewriteWarning: Module already imported so cannot be rewritten; anemoi.utils.testing
    self.import_plugin(import_spec)

-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
=========================== short test summary info ============================
FAILED test_height_level_humidity.py::test_dewpoint_to_specific_humidity_from_file
========================= 1 failed, 1 warning in 2.69s =========================

Accompanying data

No response

Organisation

DWD

[gmertes]

While it may be possible to use other shortnames that don't conflict with your local definition, the fact that the gribs used in the test are generated with a different definition (ECMWF), means that the best practice is to also use the ECMWF one when running the test to ensure there are no conflicts or unintended behaviours.

These tests also seem to target very specific variables, so I'm not sure we can even use other shortnames that don't conflict with your local definition.

The eccodes local definition path should really only be set if you know that the data you are loading is encoded with that definition, which is not the case for this pytest. Is it an option for you to just clear the local definition path before running the pytest?