Query fixes

hub/core/query/query.py

@@ -5,9 +5,10 @@
 from hub.core.tensor import Tensor
 
 
-import numpy
+import numpy as np
 
-NP_RESULT = Union[numpy.ndarray, List[numpy.ndarray]]
+
+NP_RESULT = Union[np.ndarray, List[np.ndarray]]
 NP_ACCESS = Callable[[str], NP_RESULT]
 
 
@@ -127,17 +128,17 @@ def __getitem__(self, item):
 
     @property
     def min(self):
-        """Returns numpy.min() for the tensor"""
-        return numpy.amin(self.val)
+        """Returns np.min() for the tensor"""
+        return np.amin(self.val)
 
     @property
     def max(self):
-        """Returns numpy.max() for the tensor"""
-        return numpy.amax(self.val)
+        """Returns np.max() for the tensor"""
+        return np.amax(self.val)
 
     @property
     def mean(self):
-        """Returns numpy.mean() for the tensor"""
+        """Returns np.mean() for the tensor"""
         return self.val.mean()
 
     @property
@@ -151,7 +152,13 @@ def size(self):
         return self.val.size  # type: ignore
 
     def __eq__(self, o: object) -> bool:
-        return self.val == o
+        if isinstance(self.val, (list, np.ndarray)):
+            if isinstance(o, (list, tuple)):
+                return set(o) == set(self.val)
+            else:
+                return o in self.val
+        else:
+            return self.val == o
 
     def __lt__(self, o: object) -> bool:
         return self.val < o
@@ -165,9 +172,6 @@ def __gt__(self, o: object) -> bool:
     def __ge__(self, o: object) -> bool:
         return self.val >= o
 
-    def __ne__(self, o: object) -> bool:
-        return self.val != o
-
     def __mod__(self, o: object):
         return self.val % o
 
@@ -189,6 +193,9 @@ def __mul__(self, o: object):
     def __pow__(self, o: object):
         return self.val**o
 
+    def __contains__(self, o: object):
+        return self.contains(o)
+
 
 class GroupTensor:
     def __init__(self, dataset: Dataset, wrappers, prefix: str) -> None:
@@ -228,11 +235,17 @@ def __init__(self, tensor: Tensor) -> None:
         _classes = tensor.info["class_names"]  # type: ignore
         self._classes_dict = {v: idx for idx, v in enumerate(_classes)}
 
-    def __eq__(self, o: object) -> bool:
+    def _norm_labels(self, o: object):
         if isinstance(o, str):
-            return self.val == self._classes_dict[o]
-        else:
-            return self.val == o
+            return self._classes_dict[o]
+        elif isinstance(o, int):
+            return o
+        elif isinstance(o, (list, tuple)):
+            return o.__class__(map(self._norm_labels, o))
+
+    def __eq__(self, o: object) -> bool:
+        o = self._norm_labels(o)
+        return super(ClassLabelsTensor, self).__eq__(o)
 
     def __lt__(self, o: object) -> bool:
         if isinstance(o, str):
@@ -254,8 +267,7 @@ def __ge__(self, o: object) -> bool:
             raise ValueError("label class is not comparable")
         return self.val >= o
 
-    def __ne__(self, o: object) -> bool:
-        if isinstance(o, str):
-            return self.val != self._classes_dict[o]
-        else:
-            return self.val != o
+    def contains(self, v: Any):
+        if isinstance(v, str):
+            v = self._classes_dict[v]
+        return super(ClassLabelsTensor, self).contains(v)

hub/core/query/test/test_query.py

@@ -157,3 +157,28 @@ def test_group(local_ds):
 
     result = local_ds.filter("labels.t2 == 1", progressbar=False)
     assert len(result) == 1
+
+
+def test_multi_category_labels(local_ds):
+    ds = local_ds
+    with ds:
+        ds.create_tensor("image", htype="image", sample_compression="png")
+        ds.create_tensor(
+            "label", htype="class_label", class_names=["cat", "dog", "tree"]
+        )
+        r = np.random.randint(50, 100, (32, 32, 3), dtype=np.uint8)
+        ds.image.append(r)
+        ds.label.append([0, 1])
+        ds.image.append(r + 2)
+        ds.label.append([1, 2])
+        ds.image.append(r * 2)
+        ds.label.append([0, 2])
+    view1 = ds.filter("label == 0")
+    view2 = ds.filter("label == 'cat'")
+    view3 = ds.filter("'cat' in label")
+    view4 = ds.filter("label.contains('cat')")
+    exp_images = np.array([r, r * 2])
+    exp_labels = np.array([[0, 1], [0, 2]], dtype=np.uint8)
+    for v in (view1, view2, view3, view4):
+        np.testing.assert_array_equal(v.image.numpy(), exp_images)
+        np.testing.assert_array_equal(v.label.numpy(), exp_labels)