Fixed problem when passing slices opened from NetCDF files

CHANGES.md

@@ -1,12 +1,22 @@
 ## Version 0.1.2 (in development)
 
+### Enhancements
+
 * Introduced new configuration flag `persist_mem_slices`. 
   If set, in-memory `xr.Dataset` instances will be first persisted to a 
   temporary Zarr, then reopened, and then appended to the target dataset. [#11]
-* Fixed problem where info about closing slice was logged twice. [#9]
 * Improved readability of generated configuration documentation.
 * Using `requirements-dev.txt` for development package dependencies.
 
+### Fixes
+
+* Fixed problem when passing slices opened from NetCDF files. The error was 
+  `TypeError: VariableEncoding.__init__() got an unexpected keyword argument 'chunksizes'`. 
+  [#14]
+
+* Fixed problem where info about closing slice was logged twice. [#9]
+
+
 ## Version 0.1.1
 
 Metadata fixes in `setup.cfg`. No actual code changes.

tests/test_metadata.py

@@ -238,6 +238,40 @@ def test_variable_defaults(self):
             ).to_dict(),
         )
 
+    def test_variable_encoding_from_netcdf(self):
+        a = xr.DataArray(np.zeros((2, 3, 4)), dims=("time", "y", "x"))
+        a.encoding.update(chunksizes=(16, 2, 2))  # turned into "chunks"
+        b = xr.DataArray(np.zeros((2, 3, 4)), dims=("time", "y", "x"))
+        b.encoding.update(contiguous=True, endian="big")  # logs warning
+        self.assertEqual(
+            {
+                "attrs": {},
+                "dims": {"time": 2, "x": 4, "y": 3},
+                "variables": {
+                    "a": {
+                        "attrs": {},
+                        "dims": ("time", "y", "x"),
+                        "encoding": {"chunks": (16, 2, 2)},
+                        "shape": (2, 3, 4),
+                    },
+                    "b": {
+                        "attrs": {},
+                        "dims": ("time", "y", "x"),
+                        "encoding": {},
+                        "shape": (2, 3, 4),
+                    },
+                },
+            },
+            DatasetMetadata.from_dataset(
+                xr.Dataset(
+                    {
+                        "a": a,
+                        "b": b,
+                    }
+                )
+            ).to_dict(),
+        )
+
     def test_variable_encoding_normalisation(self):
         def normalize(k, v):
             metadata = DatasetMetadata.from_dataset(

zappend/metadata.py

@@ -10,6 +10,7 @@
 import xarray as xr
 
 from .config import merge_configs, DEFAULT_APPEND_DIM
+from .log import logger
 
 
 class Undefined:
@@ -34,6 +35,7 @@ def __init__(
         calendar: str | Undefined = UNDEFINED,
         compressor: Codec | None | Undefined = UNDEFINED,
         filters: list[Codec] | None | Undefined = UNDEFINED,
+        **unknown_settings,
     ):
         """All arguments default to UNDEFINED, so they can be distinguished
         from None, which is has a special meaning for some values.
@@ -47,6 +49,8 @@ def __init__(
         self.calendar = calendar
         self.compressor = compressor
         self.filters = filters
+        if unknown_settings:
+            logger.warning("Ignoring unknown encoding settings: %s", unknown_settings)
 
     def to_dict(self):
         d = {
@@ -271,6 +275,10 @@ def _get_effective_variables(
                 encoding["fill_value"] = fill_value
         if "preferred_chunks" in encoding:
             encoding.pop("preferred_chunks")
+        if "chunksizes" in encoding:
+            chunk_sizes = encoding.pop("chunksizes")
+            if "chunks" not in encoding:
+                encoding["chunks"] = chunk_sizes
         variables[var_name] = VariableMetadata(
             dims=dims, shape=shape, encoding=VariableEncoding(**encoding), attrs=attrs
         )