Updated package to use chunk_grid in the latest delayedarray.

setup.cfg

@@ -49,7 +49,7 @@ python_requires = >=3.8
 # For more information, check out https://semver.org/.
 install_requires =
     importlib-metadata; python_version<"3.8"
-    delayedarray>=0.4.0
+    delayedarray>=0.5.0
     numpy
     h5py
 

src/hdf5array/Hdf5CompressedSparseMatrixSeed.py

@@ -1,5 +1,5 @@
 from typing import Optional, Sequence, Tuple, Callable, Literal
-from delayedarray import extract_dense_array, extract_sparse_array, chunk_shape, DelayedArray, wrap, is_sparse, SparseNdarray, is_masked
+from delayedarray import extract_dense_array, extract_sparse_array, chunk_grid, DelayedArray, wrap, is_sparse, SparseNdarray, is_masked, chunk_shape_to_grid
 from h5py import File
 import numpy
 from numpy import ndarray, dtype, integer, zeros, issubdtype, array
@@ -200,13 +200,19 @@ def is_sparse_Hdf5CompressedSparseMatrixSeed(x: Hdf5CompressedSparseMatrixSeed):
     return True
 
 
-@chunk_shape.register
-def chunk_shape_Hdf5CompressedSparseMatrixSeed(x: Hdf5CompressedSparseMatrixSeed):
-    """See :py:meth:`~delayedarray.chunk_shape.chunk_shape`."""
+@chunk_grid.register
+def chunk_grid_Hdf5CompressedSparseMatrixSeed(x: Hdf5CompressedSparseMatrixSeed):
+    """
+    See :py:meth:`~delayedarray.chunk_grid.chunk_grid`.
+
+    The cost factor is set to 20 to reflect the computational work involved in
+    extracting data from disk.
+    """
     if x._by_column:
-        return (x._shape[0], 1)
+        chunks = (x._shape[0], 1)
     else:
-        return (1, x._shape[1])
+        chunks = (1, x._shape[1])
+    return chunk_shape_to_grid(chunks, x.shape, cost_factor=20) 
 
 
 def _extract_array(

src/hdf5array/Hdf5DenseArraySeed.py

@@ -1,5 +1,5 @@
 from typing import Optional, Sequence, Tuple, Union
-from delayedarray import extract_dense_array, chunk_shape, DelayedArray, wrap, is_masked
+from delayedarray import extract_dense_array, chunk_grid, DelayedArray, wrap, is_masked, chunk_shape_to_grid
 from h5py import File
 import numpy
 from numpy import ndarray, dtype, asfortranarray, ix_
@@ -98,10 +98,15 @@ def name(self) -> str:
         return self._name
 
 
-@chunk_shape.register
-def chunk_shape_Hdf5DenseArraySeed(x: Hdf5DenseArraySeed):
-    """See :py:meth:`~delayedarray.chunk_shape.chunk_shape`."""
-    return x._chunks
+@chunk_grid.register
+def chunk_grid_Hdf5DenseArraySeed(x: Hdf5DenseArraySeed):
+    """
+    See :py:meth:`~delayedarray.chunk_grid.chunk_grid`.
+
+    The cost factor is set to 20 to reflect the computational work involved in
+    extracting data from disk.
+    """
+    return chunk_shape_to_grid(x._chunks, x._shape, cost_factor=20) 
 
 
 @extract_dense_array.register

tests/test_Hdf5CompressedSparseMatrix.py

@@ -5,6 +5,8 @@
 import tempfile
 import scipy.sparse
 
+from utils import chunk_shape
+
 __author__ = "jkanche"
 __copyright__ = "jkanche"
 __license__ = "MIT"
@@ -30,7 +32,7 @@ def test_Hdf5CompressedSparseMatrix_column():
 
     assert arr.shape == shape
     assert arr.dtype == y.dtype
-    assert delayedarray.chunk_shape(arr) == (100, 1)
+    assert chunk_shape(arr) == (100, 1)
     assert (delayedarray.to_dense_array(arr) == y.toarray()).all()
     assert not delayedarray.is_masked(arr)
 
@@ -63,7 +65,7 @@ def test_Hdf5CompressedSparseMatrix_row():
 
     assert arr.shape == shape 
     assert arr.dtype == y.dtype
-    assert delayedarray.chunk_shape(arr) == (1, 200)
+    assert chunk_shape(arr) == (1, 200)
     assert (delayedarray.to_dense_array(arr) == y.toarray()).all()
 
     # Check that consecutive slicing works as expected.
@@ -95,7 +97,7 @@ def test_Hdf5CompressedSparseMatrix_dtype():
 
     assert arr.shape == shape 
     assert arr.dtype == numpy.int16
-    assert delayedarray.chunk_shape(arr) == (55, 1)
+    assert chunk_shape(arr) == (55, 1)
 
     as_dense = delayedarray.to_dense_array(arr)
     assert (as_dense == y.toarray()).all()
@@ -136,4 +138,4 @@ def test_Hdf5CompressedSparseMatrix_to_sparse():
     assert isinstance(_to_csc, scipy.sparse.csc_matrix)
 
     _to_coo = delayedarray.to_scipy_sparse_matrix(arr, "coo")
-    assert isinstance(_to_coo, scipy.sparse.coo_matrix)
+    assert isinstance(_to_coo, scipy.sparse.coo_matrix)

tests/test_Hdf5DenseArray.py

@@ -4,6 +4,8 @@
 import delayedarray
 import tempfile
 
+from utils import chunk_shape
+
 __author__ = "jkanche"
 __copyright__ = "jkanche"
 __license__ = "MIT"
@@ -26,7 +28,7 @@ def test_Hdf5DenseArray_native():
 
     assert arr.shape == y.shape
     assert arr.dtype == y.dtype
-    assert delayedarray.chunk_shape(arr) == chunk_sizes
+    assert chunk_shape(arr) == chunk_sizes
     assert (delayedarray.to_dense_array(arr) == y).all()
     assert not delayedarray.is_masked(arr)
 
@@ -56,7 +58,7 @@ def test_Hdf5DenseArray_non_native():
     actual_chunk_sizes = (*list(reversed(chunk_sizes)),)
     assert arr.shape == actual_shape
     assert arr.dtype == y.dtype
-    assert delayedarray.chunk_shape(arr) == actual_chunk_sizes
+    assert chunk_shape(arr) == actual_chunk_sizes
     assert (delayedarray.to_dense_array(arr) == y.T).all()
 
     # Check that the slicing works as expected.
@@ -83,7 +85,7 @@ def test_Hdf5DenseArray_new_type():
 
     assert arr.shape == test_shape
     assert arr.dtype == numpy.dtype("int32")
-    assert delayedarray.chunk_shape(arr) == chunk_sizes
+    assert chunk_shape(arr) == chunk_sizes
     assert (delayedarray.to_dense_array(arr) == y.astype(numpy.int32)).all()
 
 

tests/utils.py

@@ -0,0 +1,6 @@
+import delayedarray
+
+
+def chunk_shape(x):
+    grid = delayedarray.chunk_grid(x)
+    return (grid.boundaries[0][0], grid.boundaries[1][0])