Add ndarray example previews when requested

src/pyxtend/struct.py

@@ -26,10 +26,21 @@ def struct(obj: Any, level: int = 0, limit: int = 3, examples: bool = False) ->
         # This works for both TensorFlow and PyTorch
         return {obj_type_name: [f"{obj.dtype}, shape={tuple(getattr(obj, 'shape', ()))}"]}
     elif obj_type_name == "ndarray":
-        inner_structure = "empty" if obj.size == 0 else struct(obj.item(0), level + 1)
         shape = tuple(obj.shape)
         dtype = obj.dtype.name
-        return {f"{type(obj).__name__}": [f"{dtype}, shape={shape}"]}
+        summary = f"{dtype}, shape={shape}"
+
+        if not examples:
+            return {f"{type(obj).__name__}": [summary]}
+
+        preview_limit = 3
+        flat = obj.flatten()
+        preview = flat[:preview_limit].tolist()
+
+        if flat.size > preview_limit:
+            preview.append(f"...{flat.size} total")
+
+        return {f"{type(obj).__name__}": [summary, *preview]}
     elif obj_type_name == "Polygon":
         coords = list(getattr(obj, "exterior", {}).coords) if hasattr(obj, "exterior") else []
         shape = (len(coords), len(coords[0]) if coords else 0)

tests/unit/test_struct.py

@@ -112,7 +112,15 @@ def test_numpy_array_examples():  # maybe include three?
     result = struct(np_arr, examples=True)
     # Reponse will depend on default interger type of the platform
     expected_dtype = "int32" if np.dtype("int").itemsize == 4 else "int64"
-    assert result == {"ndarray": [f"{expected_dtype}, shape=(7,)"]}
+    assert result == {"ndarray": [f"{expected_dtype}, shape=(7,)", 1, 2, 3, "...7 total"]}
+
+
+def test_numpy_array_examples_vs_structure():
+    np_arr = np.array([1, 2, 3])
+    without_examples = struct(np_arr, examples=False)
+    with_examples = struct(np_arr, examples=True)
+
+    assert without_examples != with_examples
 
 
 def test_dict():