Incompatible with iris 3.16

[bouweandela]

It looks like ncdata v0.3.2 is not compatible with the latest iris release, v3.16.0 (tested with xarray 2026.7.0).

.pixi/envs/default/lib/python3.14/site-packages/ncdata/iris_xarray.py:54: in cubes_from_xarray
    cubes = to_iris(ncdata, **iris_load_kwargs)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
.pixi/envs/default/lib/python3.14/site-packages/ncdata/iris.py:53: in to_iris
    cubes = CubeList(iris.load(dslikes, **iris_load_kwargs))
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
.pixi/envs/default/lib/python3.14/site-packages/iris/loading.py:235: in load
    cubes = _load_collection(uris, constraints, callback).combined().cubes()
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
.pixi/envs/default/lib/python3.14/site-packages/iris/loading.py:203: in _load_collection
    result = _CubeFilterCollection.from_cubes(cubes, constraints)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
.pixi/envs/default/lib/python3.14/site-packages/iris/loading.py:159: in from_cubes
    for c in cubes:
             ^^^^^
.pixi/envs/default/lib/python3.14/site-packages/iris/loading.py:108: in _generate_cubes
    for cube in iris.io.load_data_objects(data_objects, callback):
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
.pixi/envs/default/lib/python3.14/site-packages/iris/io/__init__.py:306: in load_data_objects
    yield from load_http(urls, callback)
.pixi/envs/default/lib/python3.14/site-packages/iris/io/__init__.py:286: in load_http
    for cube in handling_format_spec.handler(fnames, callback):
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
.pixi/envs/default/lib/python3.14/site-packages/iris/fileformats/netcdf/loader.py:712: in load_cubes
    with CFReader(file_source) as cf:
         ^^^^^^^^^^^^^^^^^^^^^
.pixi/envs/default/lib/python3.14/site-packages/iris/fileformats/cf.py:1356: in __init__
    ds.set_auto_chartostring(False)
    ^^^^^^^^^^^^^^^^^^^^^^^^
.pixi/envs/default/lib/python3.14/site-packages/ncdata/dataset_like.py:74: in __getattr__
    return self.getncattr(attr)
           ^^^^^^^^^^^^^^^^^^^^
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

self = <ncdata.dataset_like.Nc4DatasetLike object at 0x7c1cf981dd30>, attr = 'set_auto_chartostring'

    def getncattr(self, attr: str):
        attrs = self._ncdata.avals
        if attr in attrs:
            result = attrs[attr]
        else:
            # Don't allow it to issue a KeyError, as this upsets 'getattr' usage.
            # Raise an AttributeError instead.
>           raise AttributeError(attr)
E           AttributeError: set_auto_chartostring

.pixi/envs/default/lib/python3.14/site-packages/ncdata/dataset_like.py:59: AttributeError

I noticed you have some work in progress to fix this in #253 and #254

[bouweandela]

There also appears to be an issue with coordinates with string values when converting the other way, the code below works fine with iris 3.15.0 and crashes with 3.16.0:

In [1]: import iris.coords, iris.cube, ncdata.iris_xarray, numpy as np

In [2]: ncdata.iris_xarray.cubes_to_xarray(iris.cube.Cube(np.array([0, 1]), aux_coords_and_dims=[(iris.coords.AuxCoord(np.array(["Jan", "Feb"])), 0)]))
/home/bandela/src/esmvalgroup/esmvalcore/.pixi/envs/default/lib/python3.14/site-packages/iris/fileformats/netcdf/saver.py:2986: IrisDeprecation: Saving to netcdf with legacy-style attribute handling for backwards compatibility.
This mode is deprecated since Iris 3.8, and will eventually be removed.
Please consider enabling the new split-attributes handling mode, by setting 'iris.FUTURE.save_split_attrs = True'.
  warn_deprecated(message)
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[2], line 1
----> 1 ncdata.iris_xarray.cubes_to_xarray(iris.cube.Cube(np.array([0, 1]), aux_coords_and_dims=[(iris.coords.AuxCoord(np.array(["Jan", "Feb"])), 0)]))

File ~/src/esmvalgroup/esmvalcore/.pixi/envs/default/lib/python3.14/site-packages/ncdata/iris_xarray.py:90, in cubes_to_xarray(cubes, iris_save_kwargs, xr_load_kwargs)
     88 xr_load_kwargs = xr_load_kwargs or {}
     89 ncdata = from_iris(cubes, **iris_save_kwargs)
---> 90 xrds = to_xarray(ncdata, **xr_load_kwargs)
     91 return xrds

File ~/src/esmvalgroup/esmvalcore/.pixi/envs/default/lib/python3.14/site-packages/ncdata/xarray.py:189, in to_xarray(ncdata, **xarray_load_kwargs)
    169 def to_xarray(ncdata: NcData, **xarray_load_kwargs) -> xr.Dataset:
    170     """
    171     Convert :class:`~ncdata.NcData` to an xarray :class:`~xarray.Dataset`.
    172 
   (...)    187 
    188     """
--> 189     return _XarrayNcDataStore(ncdata).to_xarray(**xarray_load_kwargs)

File ~/src/esmvalgroup/esmvalcore/.pixi/envs/default/lib/python3.14/site-packages/ncdata/xarray.py:165, in _XarrayNcDataStore.to_xarray(self, **xr_load_kwargs)
    162 from xarray.backends.store import StoreBackendEntrypoint
    164 store_entrypoint = StoreBackendEntrypoint()
--> 165 ds = store_entrypoint.open_dataset(self, **xr_load_kwargs)
    166 return ds

File ~/src/esmvalgroup/esmvalcore/.pixi/envs/default/lib/python3.14/site-packages/xarray/backends/store.py:42, in StoreBackendEntrypoint.open_dataset(self, filename_or_obj, mask_and_scale, decode_times, concat_characters, decode_coords, drop_variables, set_indexes, use_cftime, decode_timedelta)
     27 def open_dataset(
     28     self,
     29     filename_or_obj: T_PathFileOrDataStore,
   (...)     38     decode_timedelta=None,
     39 ) -> Dataset:
     40     assert isinstance(filename_or_obj, AbstractDataStore)
---> 42     vars, attrs = filename_or_obj.load()
     43     encoding = filename_or_obj.get_encoding()
     45     vars, attrs, coord_names = conventions.decode_cf_variables(
     46         vars,
     47         attrs,
   (...)     54         decode_timedelta=decode_timedelta,
     55     )

File ~/src/esmvalgroup/esmvalcore/.pixi/envs/default/lib/python3.14/site-packages/ncdata/xarray.py:59, in _XarrayNcDataStore.load(self)
     57 variables = {}
     58 for k, v in self.ncdata.variables.items():
---> 59     xr_var = xr.Variable(
     60         v.dimensions, v.data, v.avals, getattr(v, "encoding", {})
     61     )
     62     variables[k] = xr_var
     64 return variables, self.ncdata.avals

File ~/src/esmvalgroup/esmvalcore/.pixi/envs/default/lib/python3.14/site-packages/xarray/core/variable.py:398, in Variable.__init__(self, dims, data, attrs, encoding, fastpath)
    370 def __init__(
    371     self,
    372     dims,
   (...)    376     fastpath=False,
    377 ):
    378     """
    379     Parameters
    380     ----------
   (...)    396         unrecognized encoding items.
    397     """
--> 398     super().__init__(
    399         dims=dims, data=as_compatible_data(data, fastpath=fastpath), attrs=attrs
    400     )
    402     self._encoding: dict[Any, Any] | None = None
    403     if encoding is not None:

File ~/src/esmvalgroup/esmvalcore/.pixi/envs/default/lib/python3.14/site-packages/xarray/namedarray/core.py:261, in NamedArray.__init__(self, dims, data, attrs)
    254 def __init__(
    255     self,
    256     dims: _DimsLike,
    257     data: duckarray[Any, _DType_co],
    258     attrs: _AttrsLike = None,
    259 ):
    260     self._data = data
--> 261     self._dims = self._parse_dimensions(dims)
    262     self._attrs = dict(attrs) if attrs else None

File ~/src/esmvalgroup/esmvalcore/.pixi/envs/default/lib/python3.14/site-packages/xarray/namedarray/core.py:505, in NamedArray._parse_dimensions(self, dims)
    503 dims = (dims,) if isinstance(dims, str) else tuple(dims)
    504 if len(dims) != self.ndim:
--> 505     raise ValueError(
    506         f"dimensions {dims} must have the same length as the "
    507         f"number of data dimensions, ndim={self.ndim}"
    508     )
    509 if len(set(dims)) < len(dims):
    510     repeated_dims = {d for d in dims if dims.count(d) > 1}

ValueError: dimensions ('dim0', 'string3') must have the same length as the number of data dimensions, ndim=1

[pp-mo]

Thanks @bouweandela
I also had this problem -- the set_auto_chartostring thing is pretty obvious, but there are other breakages in the ncdata tests.
I am working on it..

In fact, it looks like I personally messed up a bit adding the Iris chardata support : it did not respect the needs of ncdata + was not tested with it.

( In addition it's caused a couple of other Iris breakages,
e.g. SciTools/iris#7264 SciTools/iris#7263.
So we need to address those, and we could even be thinking of a patch release.
)