Wrap data in set_data_io with a DataChunkIterator to support overriding hdf5 dataset backend configurations

CHANGELOG.md

@@ -10,6 +10,7 @@
 
 ### Enhancements
 - Added support to append to a dataset of references for HDMF-Zarr. @mavaylon1 [#1157](https://github.com/hdmf-dev/hdmf/pull/1157)
+- Added support for overriding backend configurations of h5py.Dataset objects in Container.set_data_io: [#1172](https://github.com/hdmf-dev/hdmf/pull/1172)
 - Adjusted stacklevel of warnings to point to user code when possible. @rly [#1166](https://github.com/hdmf-dev/hdmf/pull/1166)
 - Improved "already exists" error message when adding a container to a `MultiContainerInterface`. @rly [#1165](https://github.com/hdmf-dev/hdmf/pull/1165)
 - Added support to write multidimensional string arrays. @stephprince [#1173](https://github.com/hdmf-dev/hdmf/pull/1173)

src/hdmf/container.py

@@ -2,7 +2,7 @@
 from abc import abstractmethod
 from collections import OrderedDict
 from copy import deepcopy
-from typing import Type
+from typing import Type, Optional
 from uuid import uuid4
 from warnings import warn
 import os
@@ -11,7 +11,7 @@
 import numpy as np
 import pandas as pd
 
-from .data_utils import DataIO, append_data, extend_data
+from .data_utils import DataIO, append_data, extend_data, AbstractDataChunkIterator
 from .utils import docval, get_docval, getargs, ExtenderMeta, get_data_shape, popargs, LabelledDict
 
 from .term_set import TermSet, TermSetWrapper
@@ -826,7 +826,14 @@ def __smart_str_dict(d, num_indent):
         out += '\n' + indent + right_br
         return out
 
-    def set_data_io(self, dataset_name: str, data_io_class: Type[DataIO], data_io_kwargs: dict = None, **kwargs):
+    def set_data_io(
+        self,
+        dataset_name: str,
+        data_io_class: Type[DataIO],
+        data_io_kwargs: dict = None,
+        data_chunk_iterator_class: Optional[Type[AbstractDataChunkIterator]] = None,
+        data_chunk_iterator_kwargs: dict = None, **kwargs
+    ):
         """
         Apply DataIO object to a dataset field of the Container.
 
@@ -838,9 +845,18 @@ def set_data_io(self, dataset_name: str, data_io_class: Type[DataIO], data_io_kw
             Class to use for DataIO, e.g. H5DataIO or ZarrDataIO
         data_io_kwargs: dict
             keyword arguments passed to the constructor of the DataIO class.
+        data_chunk_iterator_class: Type[AbstractDataChunkIterator]
+            Class to use for DataChunkIterator. If None, no DataChunkIterator is used.
+        data_chunk_iterator_kwargs: dict
+            keyword arguments passed to the constructor of the DataChunkIterator class.
         **kwargs:
             DEPRECATED. Use data_io_kwargs instead.
             kwargs are passed to the constructor of the DataIO class.
+
+        Notes
+        -----
+        If data_chunk_iterator_class is not None, the data is wrapped in the DataChunkIterator before being wrapped in
+        the DataIO. This allows for rewriting the backend configuration of hdf5 datasets.
         """
         if kwargs or (data_io_kwargs is None):
             warn(
@@ -851,8 +867,11 @@ def set_data_io(self, dataset_name: str, data_io_class: Type[DataIO], data_io_kw
             )
             data_io_kwargs = kwargs
         data = self.fields.get(dataset_name)
+        data_chunk_iterator_kwargs = data_chunk_iterator_kwargs or dict()
         if data is None:
             raise ValueError(f"{dataset_name} is None and cannot be wrapped in a DataIO class")
+        if data_chunk_iterator_class is not None:
+            data = data_chunk_iterator_class(data=data, **data_chunk_iterator_kwargs)
         self.fields[dataset_name] = data_io_class(data=data, **data_io_kwargs)
 
 
@@ -896,7 +915,13 @@ def set_dataio(self, **kwargs):
         dataio.data = self.__data
         self.__data = dataio
 
-    def set_data_io(self, data_io_class: Type[DataIO], data_io_kwargs: dict) -> None:
+    def set_data_io(
+        self,
+        data_io_class: Type[DataIO],
+        data_io_kwargs: dict,
+        data_chunk_iterator_class: Optional[Type[AbstractDataChunkIterator]] = None,
+        data_chunk_iterator_kwargs: dict = None,
+    ) -> None:
         """
         Apply DataIO object to the data held by this Data object.
 
@@ -906,8 +931,21 @@ def set_data_io(self, data_io_class: Type[DataIO], data_io_kwargs: dict) -> None
             The DataIO to apply to the data held by this Data.
         data_io_kwargs: dict
             The keyword arguments to pass to the DataIO.
+        data_chunk_iterator_class: Type[AbstractDataChunkIterator]
+            The DataChunkIterator to use for the DataIO. If None, no DataChunkIterator is used.
+        data_chunk_iterator_kwargs: dict
+            The keyword arguments to pass to the DataChunkIterator.
+
+        Notes
+        -----
+        If data_chunk_iterator_class is not None, the data is wrapped in the DataChunkIterator before being wrapped in
+        the DataIO. This allows for rewriting the backend configuration of hdf5 datasets.
         """
-        self.__data = data_io_class(data=self.__data, **data_io_kwargs)
+        data_chunk_iterator_kwargs = data_chunk_iterator_kwargs or dict()
+        data = self.__data
+        if data_chunk_iterator_class is not None:
+            data = data_chunk_iterator_class(data=data, **data_chunk_iterator_kwargs)
+        self.__data = data_io_class(data=data, **data_io_kwargs)
 
     @docval({'name': 'func', 'type': types.FunctionType, 'doc': 'a function to transform *data*'})
     def transform(self, **kwargs):

tests/unit/test_io_hdf5_h5tools.py

@@ -3801,6 +3801,11 @@ def __init__(self, **kwargs):
                 self.data2 = kwargs["data2"]
 
         self.obj = ContainerWithData("name", [1, 2, 3, 4, 5], None)
+        self.file_path = get_temp_filepath()
+
+    def tearDown(self):
+        if os.path.exists(self.file_path):
+            os.remove(self.file_path)
 
     def test_set_data_io(self):
         self.obj.set_data_io("data1", H5DataIO, data_io_kwargs=dict(chunks=True))
@@ -3823,6 +3828,31 @@ def test_set_data_io_old_api(self):
         self.assertIsInstance(self.obj.data1, H5DataIO)
         self.assertTrue(self.obj.data1.io_settings["chunks"])
 
+    def test_set_data_io_h5py_dataset(self):
+        file = File(self.file_path, 'w')
+        data = file.create_dataset('data', data=[1, 2, 3, 4, 5], chunks=(3,))
+        class ContainerWithData(Container):
+            __fields__ = ('data',)
+
+            @docval(
+                {"name": "name", "doc": "name", "type": str},
+                {'name': 'data', 'doc': 'field1 doc', 'type': h5py.Dataset},
+            )
+            def __init__(self, **kwargs):
+                super().__init__(name=kwargs["name"])
+                self.data = kwargs["data"]
+
+        container = ContainerWithData("name", data)
+        container.set_data_io(
+            "data",
+            H5DataIO,
+            data_io_kwargs=dict(chunks=(2,)),
+            data_chunk_iterator_class=DataChunkIterator,
+        )
+
+        self.assertIsInstance(container.data, H5DataIO)
+        self.assertEqual(container.data.io_settings["chunks"], (2,))
+        file.close()
 
 class TestDataSetDataIO(TestCase):
 
@@ -3831,8 +3861,30 @@ class MyData(Data):
             pass
 
         self.data = MyData("my_data", [1, 2, 3])
+        self.file_path = get_temp_filepath()
+
+    def tearDown(self):
+        if os.path.exists(self.file_path):
+            os.remove(self.file_path)
 
     def test_set_data_io(self):
         self.data.set_data_io(H5DataIO, dict(chunks=True))
         assert isinstance(self.data.data, H5DataIO)
         assert self.data.data.io_settings["chunks"]
+
+    def test_set_data_io_h5py_dataset(self):
+        file = File(self.file_path, 'w')
+        data = file.create_dataset('data', data=[1, 2, 3, 4, 5], chunks=(3,))
+        class MyData(Data):
+            pass
+
+        my_data = MyData("my_data", data)
+        my_data.set_data_io(
+            H5DataIO,
+            data_io_kwargs=dict(chunks=(2,)),
+            data_chunk_iterator_class=DataChunkIterator,
+        )
+
+        self.assertIsInstance(my_data.data, H5DataIO)
+        self.assertEqual(my_data.data.io_settings["chunks"], (2,))
+        file.close()