resilience-towards-missing-array-dimensions-zarr-attribute.py

resilience-towards-missing-array-dimensions-zarr-attribute.py

# %%
from pathlib import Path
import json 

import numpy as np
import xarray as xr
from datatree import DataTree, open_datatree

# %%

def load_json(path: Path) -> dict:
    with open(path, encoding="utf-8") as fp:
        return json.load(fp)


# %% [markdown]
# ## Data Creation
# 

# %%
xda = xr.DataArray(
    np.arange(3 * 18).reshape(3, 18),
    coords={"label": list("abc"), "z": list(range(18))},
)
xda = xda.chunk({"label": 2, "z": 4})
xda

# %%
xdt = DataTree(xr.Dataset({"my_xda": xda}))
xdt.my_xda

# %% [markdown]
# ## Data Writing

# %%
zarr_path = Path() / "../generated/my_array.zarr"
xdt.to_zarr(zarr_path)
print(zarr_path.resolve())

# %% [markdown]
# ## Data Initial Reading

# %%
open_datatree(zarr_path, engine="zarr").my_xda

# %% [markdown]
# ## Data Alteration

# %%
# corrupt the z's variable `_ARRAY_DIMENSIONS` xarray's attribute
zattrs_path = zarr_path / 'z/.zattrs'
assert zattrs_path.is_file()
zattrs_path.write_text('{}')

# Note: it has no impact, only the root .zmetdata seems to be used

# %%
open_datatree(zarr_path, engine="zarr")

# %%
zattrs_path = zarr_path / 'label/.zattrs'
assert zattrs_path.is_file()
zattrs_path.write_text('{}')

# %%
open_datatree(zarr_path, engine="zarr")

# %%
zattrs_path = zarr_path / 'my_xda/.zattrs'
assert zattrs_path.is_file()
zattrs_path.write_text('{}')

# %%
open_datatree(zarr_path, engine="zarr")

# %%
zmetadata_path = zarr_path / '.zmetadata'
assert zmetadata_path.is_file()
zmetadata=load_json(zmetadata_path)
zmetadata['metadata']['z/.zattrs'] = {}
zmetadata_path.write_text(json.dumps(zmetadata, indent=4))

# %%
open_datatree(zarr_path, engine="zarr")