fix casting bug (#3525)

apis/python/src/tiledbsoma/_indexer.py

@@ -74,12 +74,12 @@ def __init__(
 
         # TODO: the map_locations interface does not accept chunked arrays. It would
         # save a copy (reduce memory usage) if they were natively supported.
-        if isinstance(
-            data, (pa.Array, pa.ChunkedArray, pd.arrays.IntegerArray, pd.Series)
-        ):
+        if isinstance(data, (pa.Array, pa.ChunkedArray)):
             data = data.to_numpy()
         elif isinstance(data, list):
             data = np.array(data, dtype=np.int64)
+        elif isinstance(data, (pd.arrays.IntegerArray, pd.Series)):
+            data = data.to_numpy(dtype=np.int64, copy=False)
 
         self._reindexer.map_locations(data)
 
@@ -95,7 +95,7 @@ def get_indexer(self, target: IndexerDataType) -> npt.NDArray[np.intp]:
             return self._reindexer.get_indexer_pyarrow(target)
 
         if isinstance(target, (pd.arrays.IntegerArray, pd.Series)):
-            target = target.to_numpy()
+            target = target.to_numpy(dtype=np.int64, copy=False)
         elif isinstance(target, list):
             target = np.array(target, dtype=np.int64)
 

apis/python/tests/test_indexer.py

@@ -102,9 +102,12 @@ def test_indexer(contextual: bool, keys: np.array, lookups: np.array):
     num_threads = 10
 
     def target():
-        indexer = IntIndexer(keys, context=context)
-        results = indexer.get_indexer(lookups)
-        all_results.append(results)
+        try:
+            indexer = IntIndexer(keys, context=context)
+            results = indexer.get_indexer(lookups)
+            all_results.append(results)
+        except Exception as e:
+            all_results.append(e)
 
     for t in range(num_threads):
         thread = threading.Thread(target=target, args=())
@@ -113,6 +116,8 @@ def target():
     panda_indexer = pd.Index(keys)
     panda_results = panda_indexer.get_indexer(lookups)
     for i in range(num_threads):
+        if isinstance(all_results[i], Exception):
+            raise all_results[i]
         np.testing.assert_equal(all_results[i].all(), panda_results.all())