Fix numeric slicing for compound dtype datasets

lindi/LindiH5pyFile/LindiH5pyDataset.py

@@ -214,9 +214,22 @@ def _get_item_for_zarr(self, zarr_array: zarr.Array, selection: Any):
                 )
                 return ret
             else:
-                raise TypeError(
-                    f"Compound dataset {self.name} does not support selection with {selection}"
-                )
+                # Numeric slicing (e.g., [:], [0], [0:2])
+                # Get the raw data from zarr (will be object array of lists)
+                raw_data = zarr_array[selection]
+                # Convert to structured array with compound dtype
+                if isinstance(raw_data, np.ndarray):
+                    # Multiple elements - convert each list to tuple
+                    tuples = [tuple(item) if isinstance(item, (list, np.ndarray)) else item for item in raw_data]
+                    return np.array(tuples, dtype=self._compound_dtype)
+                elif isinstance(raw_data, (list, tuple)):
+                    # Single element - raw_data is a list or tuple
+                    return np.array(tuple(raw_data), dtype=self._compound_dtype)[()]
+                else:
+                    # Unexpected type - this shouldn't happen, but handle gracefully
+                    raise TypeError(
+                        f"Unexpected data type from zarr array for compound dataset {self.name}: {type(raw_data)}"
+                    )
 
         # We use zarr's slicing, except in the case of a scalar dataset
         if self.ndim == 0:

tests/test_lindi_h5py_file.py

@@ -2,6 +2,8 @@
 import os
 import pytest
 import h5py
+import numpy as np
+import zarr
 import lindi
 from .utils import assert_h5py_files_equal
 
@@ -357,6 +359,57 @@ def test_create_dataset():
             assert ds.shape == (3,)
 
 
+def test_compound_dtype_slicing():
+    """Test that compound dtype datasets support numeric slicing (e.g., [:], [0], [0:2])"""
+
+    with tempfile.TemporaryDirectory() as tmpdir:
+        dirname = f'{tmpdir}/test.zarr'
+        store = zarr.DirectoryStore(dirname)
+        zarr.group(store=store)
+
+        # Write compound dataset
+        with lindi.LindiH5pyFile.from_zarr_store(store, mode='r+') as h5f:
+            compound_dtype = np.dtype([('x', np.int32), ('y', np.float64)])
+            data = np.array([(1, 2.2), (3, 4.4), (5, 6.6)], dtype=compound_dtype)
+            h5f.create_dataset('dset_compound', data=data)
+
+        # Read and test slicing
+        with lindi.LindiH5pyFile.from_zarr_store(store, mode='r') as h5f:
+            dset = h5f['dset_compound']
+
+            # Test full slice
+            result = dset[:]
+            assert result.shape == (3,)
+            assert result.dtype == compound_dtype
+            assert result[0]['x'] == 1
+            assert result[0]['y'] == 2.2
+            assert result[1]['x'] == 3
+            assert result[1]['y'] == 4.4
+
+            # Test single element access
+            result = dset[0]
+            assert isinstance(result, np.void)
+            assert result['x'] == 1
+            assert result['y'] == 2.2
+
+            # Test partial slice
+            result = dset[0:2]
+            assert result.shape == (2,)
+            assert result.dtype == compound_dtype
+            assert result[0]['x'] == 1
+            assert result[1]['x'] == 3
+
+            # Test negative indexing
+            result = dset[-1]
+            assert result['x'] == 5
+            assert result['y'] == 6.6
+
+            # Test field selection (should still work)
+            result = dset['x'][:]
+            assert result.shape == (3,)
+            assert np.array_equal(result, [1, 3, 5])
+
+
 def create_example_h5_file(fname):
     with h5py.File(fname, 'w') as f:
         f.attrs['attr1'] = 'value1'