Get core Dask functionality from cfdm

Changelog.rst

@@ -1,7 +1,7 @@
 version NEXTVERSION
 -------------------
 
-**2024-??-??**
+**2024-12-??**
 
 * Allow ``'nearest_dtos'`` 2-d regridding to work with discrete
   sampling geometry source grids
@@ -23,6 +23,8 @@ version NEXTVERSION
 * New class `cf.NetCDF4Array`
 * New class `cf.CFAH5netcdfArray`
 * New class `cf.CFANetCDF4Array`
+* Replace core `dask` functionality with that imported from `cfdm`
+  (https://github.com/NCAS-CMS/cf-python/pull/836)
 * Fix bug that sometimes puts an incorrect ``radian-1`` or
   ``radian-2`` in the returned units of the differential operator
   methods and functions

cf/constants.py

@@ -37,7 +37,7 @@
       Find the total amount of physical memory (in bytes).
 
     CHUNKSIZE: `int`
-      The chunk size (in bytes) for data storage and processing.
+      The Dask chunk size (in bytes). See `cf.chunksize`.
 
     TEMPDIR: `str`
       The location to store temporary files. By default it is the

cf/data/array/fullarray.py

@@ -1,8 +1,8 @@
 import numpy as np
+from cfdm.data.mixin import IndexMixin
 
 from ...functions import indices_shape, parse_indices
 from .abstract import Array
-from .mixin import IndexMixin
 
 _FULLARRAY_HANDLED_FUNCTIONS = {}
 

cf/data/array/h5netcdfarray.py

@@ -1,13 +1,11 @@
 import cfdm
 
 from ...mixin_container import Container
-from .locks import netcdf_lock
-from .mixin import ActiveStorageMixin, ArrayMixin, FileArrayMixin, IndexMixin
+from .mixin import ActiveStorageMixin, ArrayMixin, FileArrayMixin
 
 
 class H5netcdfArray(
     ActiveStorageMixin,
-    IndexMixin,
     FileArrayMixin,
     ArrayMixin,
     Container,
@@ -23,59 +21,3 @@ class H5netcdfArray(
     .. versionadded:: NEXTVERSION
 
     """
-
-    def __dask_tokenize__(self):
-        """Return a value fully representative of the object.
-
-        .. versionadded:: NEXTVERSION
-
-        """
-        return super().__dask_tokenize__() + (self.get_mask(),)
-
-    @property
-    def _lock(self):
-        """Set the lock for use in `dask.array.from_array`.
-
-        Returns a lock object because concurrent reads are not
-        currently supported by the HDF5 library. The lock object will
-        be the same for all `NetCDF4Array` and `H5netcdfArray`
-        instances, regardless of the dataset they access, which means
-        that access to all netCDF and HDF files coordinates around the
-        same lock.
-
-        .. versionadded:: NEXTVERSION
-
-        """
-        return netcdf_lock
-
-    def _get_array(self, index=None):
-        """Returns a subspace of the dataset variable.
-
-        .. versionadded:: NEXTVERSION
-
-        .. seealso:: `__array__`, `index`
-
-        :Parameters:
-
-            {{index: `tuple` or `None`, optional}}
-
-        :Returns:
-
-            `numpy.ndarray`
-                The subspace.
-
-        """
-        if index is None:
-            index = self.index()
-
-        # We need to lock because the netCDF file is about to be accessed.
-        self._lock.acquire()
-
-        # It's cfdm.H5netcdfArray.__getitem__ that we want to
-        # call here, but we use 'Container' in super because
-        # that comes immediately before cfdm.H5netcdfArray in
-        # the method resolution order.
-        array = super(Container, self).__getitem__(index)
-
-        self._lock.release()
-        return array

cf/data/array/mixin/__init__.py

@@ -3,4 +3,3 @@
 from .cfamixin import CFAMixin
 from .compressedarraymixin import CompressedArrayMixin
 from .filearraymixin import FileArrayMixin
-from .indexmixin import IndexMixin

cf/data/array/mixin/cfamixin.py

@@ -3,8 +3,7 @@
 from itertools import accumulate, product
 
 import numpy as np
-
-from ...utils import chunk_locations, chunk_positions
+from cfdm.data.utils import chunk_locations, chunk_positions
 
 
 class CFAMixin:

cf/data/array/mixin/compressedarraymixin.py

@@ -76,12 +76,11 @@ def to_dask_array(self, chunks="auto"):
         from functools import partial
 
         import dask.array as da
+        from cfdm.data.utils import normalize_chunks
         from dask import config
         from dask.array.core import getter
         from dask.base import tokenize
 
-        from ...utils import normalize_chunks
-
         name = (f"{self.__class__.__name__}-{tokenize(self)}",)
 
         dtype = self.dtype