Append fails with shape error

Hi,

I am trying to append json references to a parquet store with

MultiZarrToZarr.append(
    [json],
    out,
    coo_map={"time":"cf:time"},
    concat_dims=['time'],
    identical_dims=["lat","lon","plev","value"],
    remote_protocol="file",
).translate()

and I ran into the issue that the group.create_dataset here cannot set new data to z["time"] becaues zarr expects another shape: Exactly the shape that already exists in the parquet store out.

ValueError                                Traceback (most recent call last)
Cell In[21], line 12
      1 MultiZarrToZarr.append(
      2     dflist[11:13],
      3     out,
      4     coo_map={"time":"cf:time"},
      5     concat_dims=['time'],
      6     identical_dims=["lat","lon","plev","value"],
      7     preprocess=None,
      8     remote_protocol="file",
      9     #remote_options=dict(lazy=True),
     10     #coo_dtypes={"time": "int64"},
     11     #postprocess=postprocess
---> 12 ).translate()

File ~/kerchunk/kerchunk/combine.py:604, in MultiZarrToZarr.translate(self, filename, storage_options)
    602     self.first_pass()
    603 if 2 not in self.done:
--> 604     self.store_coords()
    605 if 3 not in self.done:
    606     self.second_pass()

File ~/kerchunk/kerchunk/combine.py:417, in MultiZarrToZarr.store_coords(self)
    413     elif k in z:
    414         # Fall back to existing fill value
    415         kw["fill_value"] = z[k].fill_value
--> 417 arr = group.create_dataset(
    418     name=k,
    419     data=data,
    420     overwrite=True,
    421     compressor=compression,
    422     dtype=self.coo_dtypes.get(k, data.dtype),
    423     **kw,
    424 )
    425 if k in z:
    426     # copy attributes if values came from an original variable
    427     arr.attrs.update(z[k].attrs)

File /work/bm0021/conda-envs/xesmf/lib/python3.10/site-packages/zarr/hierarchy.py:1094, in Group.create_dataset(self, name, **kwargs)
   1035 """Create an array.
   1036 
   1037 Arrays are known as "datasets" in HDF5 terminology. For compatibility
   (...)
   1090 
   1091 """
   1092 assert "mode" not in kwargs
-> 1094 return self._write_op(self._create_dataset_nosync, name, **kwargs)

File /work/bm0021/conda-envs/xesmf/lib/python3.10/site-packages/zarr/hierarchy.py:935, in Group._write_op(self, f, *args, **kwargs)
    932     lock = self._synchronizer[group_meta_key]
    934 with lock:
--> 935     return f(*args, **kwargs)

File /work/bm0021/conda-envs/xesmf/lib/python3.10/site-packages/zarr/hierarchy.py:1110, in Group._create_dataset_nosync(self, name, data, **kwargs)
   1107     a = create(store=self._store, path=path, chunk_store=self._chunk_store, **kwargs)
   1109 else:
-> 1110     a = array(data, store=self._store, path=path, chunk_store=self._chunk_store, **kwargs)
   1112 return a

File /work/bm0021/conda-envs/xesmf/lib/python3.10/site-packages/zarr/creation.py:439, in array(data, **kwargs)
    436 z = create(**kwargs)
    438 # fill with data
--> 439 z[...] = data
    441 # set read_only property afterwards
    442 z.read_only = read_only

File /work/bm0021/conda-envs/xesmf/lib/python3.10/site-packages/zarr/core.py:1497, in Array.__setitem__(self, selection, value)
   1495     self.set_orthogonal_selection(pure_selection, value, fields=fields)
   1496 else:
-> 1497     self.set_basic_selection(pure_selection, value, fields=fields)

File /work/bm0021/conda-envs/xesmf/lib/python3.10/site-packages/zarr/core.py:1593, in Array.set_basic_selection(self, selection, value, fields)
   1591     return self._set_basic_selection_zd(selection, value, fields=fields)
   1592 else:
-> 1593     return self._set_basic_selection_nd(selection, value, fields=fields)

File /work/bm0021/conda-envs/xesmf/lib/python3.10/site-packages/zarr/core.py:1983, in Array._set_basic_selection_nd(self, selection, value, fields)
   1977 def _set_basic_selection_nd(self, selection, value, fields=None):
   1978     # implementation of __setitem__ for array with at least one dimension
   1979 
   1980     # setup indexer
   1981     indexer = BasicIndexer(selection, self)
-> 1983     self._set_selection(indexer, value, fields=fields)

File /work/bm0021/conda-envs/xesmf/lib/python3.10/site-packages/zarr/core.py:2011, in Array._set_selection(self, indexer, value, fields)
   2009     if not hasattr(value, "shape"):
   2010         value = np.asanyarray(value, like=self._meta_array)
-> 2011     check_array_shape("value", value, sel_shape)
   2013 # iterate over chunks in range
   2014 if (
   2015     not hasattr(self.chunk_store, "setitems")
   2016     or self._synchronizer is not None
   2017     or any(map(lambda x: x == 0, self.shape))
   2018 ):
   2019     # iterative approach

File /work/bm0021/conda-envs/xesmf/lib/python3.10/site-packages/zarr/util.py:561, in check_array_shape(param, array, shape)
    557     raise TypeError(
    558         "parameter {!r}: expected an array-like object, got {!r}".format(param, type(array))
    559     )
    560 if array.shape != shape:
--> 561     raise ValueError(
    562         "parameter {!r}: expected array with shape {!r}, got {!r}".format(
    563             param, shape, array.shape
    564         )
    565     )

ValueError: parameter 'value': expected array with shape (87665,), got (105209,)

I was trying to find a work around but I do not know how to prevent zarr from checking this.

Read kerchunk/combine.py around store_coords and the MultiZarrToZarr.append/translate flow. Reproduce the provided call against the existing parquet store, then inspect how z["time"] is shaped before group.create_dataset. Done means the append no longer raises the reported Zarr shape error for this case.