fix array casting for dtypes that have a shape attr but nothing in it

nwb_linkml/src/nwb_linkml/maps/hdf5.py

@@ -23,7 +23,6 @@ from pydantic import BaseModel, ConfigDict, Field
 
 from nwb_linkml.annotations import unwrap_optional
 from nwb_linkml.maps import Map
-from nwb_linkml.maps.hdmf import dynamictable_to_model
 from nwb_linkml.types.hdf5 import HDF5_Path
 
 if sys.version_info.minor >= 11:
@@ -234,63 +233,64 @@ class PruneEmpty(HDF5Map):
         return H5ReadResult.model_construct(path=src.path, source=src, completed=True)
 
 
-class ResolveDynamicTable(HDF5Map):
+#
-    """
+# class ResolveDynamicTable(HDF5Map):
-    Handle loading a dynamic table!
+#     """
-
+#     Handle loading a dynamic table!
-    Dynamic tables are sort of odd in that their models don't include their fields
+#
-    (except as a list of strings in ``colnames`` ),
+#     Dynamic tables are sort of odd in that their models don't include their fields
-    so we need to create a new model that includes fields for each column,
+#     (except as a list of strings in ``colnames`` ),
-    and then we include the datasets as :class:`~numpydantic.interface.hdf5.H5ArrayPath`
+#     so we need to create a new model that includes fields for each column,
-    objects which lazy load the arrays in a thread/process safe way.
+#     and then we include the datasets as :class:`~numpydantic.interface.hdf5.H5ArrayPath`
-
+#     objects which lazy load the arrays in a thread/process safe way.
-    This map also resolves the child elements,
+#
-    indicating so by the ``completes`` field in the :class:`.ReadResult`
+#     This map also resolves the child elements,
-    """
+#     indicating so by the ``completes`` field in the :class:`.ReadResult`
-
+#     """
-    phase = ReadPhases.read
+#
-    priority = 1
+#     phase = ReadPhases.read
-
+#     priority = 1
-    @classmethod
+#
-    def check(
+#     @classmethod
-        cls, src: H5SourceItem, provider: "SchemaProvider", completed: Dict[str, H5ReadResult]
+#     def check(
-    ) -> bool:
+#         cls, src: H5SourceItem, provider: "SchemaProvider", completed: Dict[str, H5ReadResult]
-        if src.h5_type == "dataset":
+#     ) -> bool:
-            return False
+#         if src.h5_type == "dataset":
-        if "neurodata_type" in src.attrs:
+#             return False
-            if src.attrs["neurodata_type"] == "DynamicTable":
+#         if "neurodata_type" in src.attrs:
-                return True
+#             if src.attrs["neurodata_type"] == "DynamicTable":
-            # otherwise, see if it's a subclass
+#                 return True
-            model = provider.get_class(src.attrs["namespace"], src.attrs["neurodata_type"])
+#             # otherwise, see if it's a subclass
-            # just inspect the MRO as strings rather than trying to check subclasses because
+#             model = provider.get_class(src.attrs["namespace"], src.attrs["neurodata_type"])
-            # we might replace DynamicTable in the future, and there isn't a stable DynamicTable
+#             # just inspect the MRO as strings rather than trying to check subclasses because
-            # class to inherit from anyway because of the whole multiple versions thing
+#             # we might replace DynamicTable in the future, and there isn't a stable DynamicTable
-            parents = [parent.__name__ for parent in model.__mro__]
+#             # class to inherit from anyway because of the whole multiple versions thing
-            return "DynamicTable" in parents
+#             parents = [parent.__name__ for parent in model.__mro__]
-        else:
+#             return "DynamicTable" in parents
-            return False
+#         else:
-
+#             return False
-    @classmethod
+#
-    def apply(
+#     @classmethod
-        cls, src: H5SourceItem, provider: "SchemaProvider", completed: Dict[str, H5ReadResult]
+#     def apply(
-    ) -> H5ReadResult:
+#         cls, src: H5SourceItem, provider: "SchemaProvider", completed: Dict[str, H5ReadResult]
-        with h5py.File(src.h5f_path, "r") as h5f:
+#     ) -> H5ReadResult:
-            obj = h5f.get(src.path)
+#         with h5py.File(src.h5f_path, "r") as h5f:
-
+#             obj = h5f.get(src.path)
-            # make a populated model :)
+#
-            base_model = provider.get_class(src.namespace, src.neurodata_type)
+#             # make a populated model :)
-            model = dynamictable_to_model(obj, base=base_model)
+#             base_model = provider.get_class(src.namespace, src.neurodata_type)
-
+#             model = dynamictable_to_model(obj, base=base_model)
-            completes = [HDF5_Path(child.name) for child in obj.values()]
+#
-
+#             completes = [HDF5_Path(child.name) for child in obj.values()]
-        return H5ReadResult(
+#
-            path=src.path,
+#         return H5ReadResult(
-            source=src,
+#             path=src.path,
-            result=model,
+#             source=src,
-            completes=completes,
+#             result=model,
-            completed=True,
+#             completes=completes,
-            applied=["ResolveDynamicTable"],
+#             completed=True,
-        )
+#             applied=["ResolveDynamicTable"],
+#         )
 
 
 class ResolveModelGroup(HDF5Map):

nwb_linkml/src/nwb_linkml/models/pydantic/hdmf_common/v1_8_0/hdmf_common_table.py

@@ -1,12 +1,7 @@
 from __future__ import annotations
-from datetime import datetime, date
+from ...hdmf_common.v1_8_0.hdmf_common_base import Data
-from decimal import Decimal
-from enum import Enum
-import re
-import sys
-from ...hdmf_common.v1_8_0.hdmf_common_base import Data, Container
 from pandas import DataFrame, Series
-from typing import Any, ClassVar, List, Literal, Dict, Optional, Union, overload, Tuple
+from typing import Any, ClassVar, List, Dict, Optional, Union, overload, Tuple
 from pydantic import (
     BaseModel,
     ConfigDict,
@@ -282,7 +277,7 @@ class DynamicTableMixin(BaseModel):
                 # special case where pandas will unpack a pydantic model
                 # into {n_fields} rows, rather than keeping it in a dict
                 val = Series([val])
-            elif isinstance(rows, int) and hasattr(val, "shape") and len(val) > 1:
+            elif isinstance(rows, int) and hasattr(val, "shape") and val.shape and val.shape[0] > 1:
                 # special case where we are returning a row in a ragged array,
                 # same as above - prevent pandas pivoting to long
                 val = Series([val])

nwb_linkml/tests/test_maps/test_hdmf.py

@@ -1,24 +0,0 @@
-import time
-
-import h5py
-import pytest
-
-from nwb_linkml.maps.hdmf import dynamictable_to_model, model_from_dynamictable
-
-NWBFILE = "/Users/jonny/Dropbox/lab/p2p_ld/data/nwb/sub-738651046_ses-760693773.nwb"
-
-
-@pytest.mark.xfail()
-@pytest.mark.parametrize("dataset", ["aibs.nwb"])
-def test_make_dynamictable(data_dir, dataset):
-    nwbfile = data_dir / dataset
-    h5f = h5py.File(nwbfile, "r")
-    group = h5f["units"]
-
-    start_time = time.time()
-    model = model_from_dynamictable(group)
-    data = dynamictable_to_model(group, model)
-
-    _ = data.model_dump_json()
-    end_time = time.time()
-    total_time = end_time - start_time