dev

Author: davidhassell

Changelog.rst

@@ -1,7 +1,7 @@
 Version NEXTVERSION
 --------------
 
-**2025-??-??**
+**2026-??-??**
 
 * Support for HEALPix grids
   (https://github.com/NCAS-CMS/cf-python/issues/909)
@@ -17,11 +17,23 @@ Version NEXTVERSION
   (https://github.com/NCAS-CMS/cf-python/issues/909)
 * New function: `cf.locate`
   (https://github.com/NCAS-CMS/cf-python/issues/909)
-* Reduce the time taken to import `cf`
-  (https://github.com/NCAS-CMS/cf-python/issues/902)
 * New optional dependency: ``healpix>=2025.1``
 * Changed dependency: ``cfdm>=1.13.0.0, <1.13.1.0``
 
+Version 3.18.3
+--------------
+
+**2026-01-??**
+
+* Write Zarr v3 datasets with `cf.write`, and allow the reading of
+  grouped Zarr v2 and v3 datasets with `cf.read`
+  (https://github.com/NCAS-CMS/cf-python/issues/895)
+* Read Zarr v2 and v3 datasets that contain a group hierarchy with
+  `cf.read` (https://github.com/NCAS-CMS/cf-python/issues/894)
+* Reduce the time taken to import `cf`
+  (https://github.com/NCAS-CMS/cf-python/issues/902)
+* New optional dependency: ``zarr>=3.1.3``
+
 ----
 
 Version 3.18.2

README.md

@@ -87,7 +87,8 @@ of its array manipulation and can:
 * read field constructs from netCDF, CDL, Zarr, PP and UM datasets with a
   choice of netCDF backends,and in local, http, and s3 locations,
 * create new field constructs in memory,
-* write and append field and domain constructs to netCDF datasets on disk,
+* write and append field and domain constructs to netCDF and Zarr v3
+  datasets on disk,
 * read, create, and manipulate UGRID mesh topologies,
 * read, write, and manipulate HEALPix grids,
 * read, write, and create coordinates defined by geometry cells,

cf/field.py

@@ -5166,7 +5166,7 @@ def healpix_decrease_refinement_level(
         )
 
         # ------------------------------------------------------------
-        # Chenge the refinement level of the Field's data
+        # Change the refinement level of the Field's data
         # ------------------------------------------------------------
 
         # Note: Using 'Data.coarsen' works because a) the HEALPix

cf/functions.py

@@ -3184,7 +3184,7 @@ def environment(display=True, paths=True):
     netCDF4: 1.7.2 /home/miniconda3/lib/python3.12/site-packages/netCDF4/__init__.py
     h5netcdf: 1.3.0 /home/miniconda3/lib/python3.12/site-packages/h5netcdf/__init__.py
     h5py: 3.12.1 /home/miniconda3/lib/python3.12/site-packages/h5py/__init__.py
-    zarr: 3.0.8 /home/miniconda3/lib/python3.12/site-packages/zarr/__init__.py
+    zarr: 3.1.3 /home/miniconda3/lib/python3.12/site-packages/zarr/__init__.py
     s3fs: 2024.12.0 /home/miniconda3/lib/python3.12/site-packages/s3fs/__init__.py
     scipy: 1.15.1 /home/miniconda3/lib/python3.12/site-packages/scipy/__init__.py
     dask: 2025.5.1 /home/miniconda3/lib/python3.12/site-packages/dask/__init__.py
@@ -3210,7 +3210,7 @@ def environment(display=True, paths=True):
     netCDF4: 1.7.2
     h5netcdf: 1.3.0
     h5py: 3.12.1
-    zarr: 3.0.8
+    zarr: 3.1.3
     s3fs: 2024.12.0
     scipy: 1.15.1
     dask: 2025.5.1

cf/mixin/fielddomain.py

@@ -2328,7 +2328,7 @@ def healpix_to_ugrid(self, cache=True, inplace=False):
             )
 
         # Create the UGRID Domain Topology construct, by creating a
-        # unique integer identifer for each unique node location.
+        # unique integer identifier for each unique node location.
         bounds_y = bounds_y.data.to_dask_array(_force_mask_hardness=False)
         bounds_x = bounds_x.data.to_dask_array(_force_mask_hardness=False)
 
@@ -2466,15 +2466,15 @@ def create_latlon_coordinates(
         # See if any lat/lon coordinates could be created
         # ------------------------------------------------------------
 
-        # See if there are any existing latitude/longutude coordinates
+        # See if there are any existing latitude/longitude coordinates
         latlon_coordinates = {
             key: c
             for key, c in f.coordinates(todict=True).items()
             if c.Units.islatitude or c.Units.islongitude
         }
         if latlon_coordinates:
             if overwrite:
-                # Remove existing latitude/longutude coordinates
+                # Remove existing latitude/longitude coordinates
                 # before carrying on
                 for key in latlon_coordinates:
                     f.del_construct(key)
@@ -2542,7 +2542,7 @@ def create_latlon_coordinates(
         # coordinates.
         # ------------------------------------------------------------
 
-        # Initialize the flag that tells us if any new coordinates
+        # Initialise the flag that tells us if any new coordinates
         # have been created
         coords_created = False
 

cf/read_write/read.py

@@ -316,6 +316,10 @@ class read(cfdm.read):
 
             .. versionadded:: 3.17.0
 
+        {{read store_dataset_shards: `bool`, optional}}
+
+            .. versionadded:: NEXTVERSION
+
         {{read cfa: `dict`, optional}}
 
             .. versionadded:: 3.15.0
@@ -328,6 +332,10 @@ class read(cfdm.read):
 
             .. versionadded:: 3.17.0
 
+        {{read group_dimension_search: `str`, optional}}
+
+            .. versionadded:: (cfdm) NEXTVERSION
+
         umversion: deprecated at version 3.0.0
             Use the *um* parameter instead.
 
@@ -434,6 +442,7 @@ def __new__(
         warn_valid=False,
         dask_chunks="storage-aligned",
         store_dataset_chunks=True,
+        store_dataset_shards=True,
         domain=False,
         cfa=None,
         cfa_write=None,
@@ -445,6 +454,7 @@ def __new__(
         ignore_read_error=False,
         fmt=None,
         file_type=None,
+        group_dimension_search="closest_ancestor",
     ):
         """Read field or domain constructs from a dataset."""
         kwargs = locals()

cf/read_write/um/umread.py

@@ -3575,7 +3575,7 @@ def read(
                 # Return now if there are valid file types
                 return []
 
-        f = self.file_open(filename, parse=True)
+        f = self.dataset_open(filename, parse=True)
 
         info = is_log_level_info(logger)
 
@@ -3598,7 +3598,7 @@ def read(
             for var in f.vars
         ]
 
-        self.file_close()
+        self.dataset_close()
 
         return [field for x in um for field in x.fields if field]
 
@@ -3632,7 +3632,7 @@ def _open_um_file(
                 The open PP or FF file object.
 
         """
-        self.file_close()
+        self.dataset_close()
         try:
             f = File(
                 filename,
@@ -3678,15 +3678,15 @@ def is_um_file(self, filename):
         try:
             # Note: No need to completely parse the file to ascertain
             #       if it's PP or FF.
-            self.file_open(filename, parse=False)
+            self.dataset_open(filename, parse=False)
         except Exception:
-            self.file_close()
+            self.dataset_close()
             return False
         else:
-            self.file_close()
+            self.dataset_close()
             return True
 
-    def file_close(self):
+    def dataset_close(self):
         """Close the file that has been read.
 
         :Returns:
@@ -3700,7 +3700,7 @@ def file_close(self):
 
         self._um_file = None
 
-    def file_open(self, filename, parse=True):
+    def dataset_open(self, filename, parse=True):
         """Open the file for reading.
 
         :Paramters:

cf/test/test_Data.py

@@ -4717,12 +4717,12 @@ def test_Data_coarsen(self):
         self.assertEqual(e.shape, (2, 2))
         self.assertTrue((e.array == [[0, 3], [12, 15]]).all())
 
-        # Non-full caorsening neighbourhood with trim_excess=True
+        # Non-full coarsening neighbourhood with trim_excess=True
         e = d.coarsen(np.max, {-1: 5}, trim_excess=True)
         self.assertEqual(e.shape, (4, 1))
         self.assertTrue((e.array == [[4], [10], [16], [22]]).all())
 
-        # Non-full caorsening neighbourhood with trim_excess=False
+        # Non-full coarsening neighbourhood with trim_excess=False
         with self.assertRaises(ValueError):
             d.coarsen(np.max, {-1: 5})
 

cf/test/test_functions.py

@@ -451,7 +451,7 @@ def test_locate(self):
             with self.assertRaises(ValueError):
                 cf.locate(lat, 30, f)
 
-        # Invalid grid types (regulat lat/lon, geometry, UGRID)
+        # Invalid grid types (regular lat/lon, geometry, UGRID)
         for f in cf.example_fields(0, 6, 8):
             with self.assertRaises(ValueError):
                 cf.locate(60, 30, f)

cf/test/test_regrid_healpix.py

@@ -56,7 +56,7 @@ def setUp(self):
 
     @unittest.skipUnless(esmpy_imported, "Requires esmpy package.")
     def test_Field_regrid_mesh_to_healpix(self):
-        # Check that UGRID -> healpix is the same as UGRID -> UGRUD
+        # Check that UGRID -> healpix is the same as UGRID -> UGRID
         self.assertFalse(cf.regrid_logging())
 
         dst = cf.Domain.create_healpix(3)  # 768 cells
@@ -85,7 +85,7 @@ def test_Field_regrid_mesh_to_healpix(self):
 
     @unittest.skipUnless(esmpy_imported, "Requires esmpy package.")
     def test_Field_regrid_healpix_to_mesh(self):
-        # Check that healpix -> UGRID is the same as UGRID -> UGRUD
+        # Check that healpix -> UGRID is the same as UGRID -> UGRID
         self.assertFalse(cf.regrid_logging())
 
         src = cf.Field(source=cf.Domain.create_healpix(3))  # 768 cells