[AL-2335] Fixes to .data to make returned results identical to python

deeplake/core/dataset/deeplake_query_dataset.py

@@ -158,7 +158,7 @@ def __getitem__(
                             "Indexing by integer in a for loop, like `for i in range(len(ds)): ... ds[i]` can be quite slow. Use `for i, sample in enumerate(ds)` instead."
                         )
                 ret = DeepLakeQueryDataset(
-                    deeplake_ds=self.deeplake_ds,
+                    deeplake_ds=self.deeplake_ds[item],
                     indra_ds=self.indra_ds[item],
                 )
         else:
@@ -305,7 +305,7 @@ def copy(self, *args, **kwargs):
         )
 
     def __del__(self):
-        self.indra_ds = None
+        pass
 
     def random_split(self, lengths: Sequence[Union[int, float]]):
         if math.isclose(sum(lengths), 1) and sum(lengths) <= 1:

deeplake/core/dataset/deeplake_query_tensor.py

@@ -74,15 +74,15 @@ def numpy(
 
     def text(self, fetch_chunks: bool = False):
         """Return text data. Only applicable for tensors with 'text' base htype."""
-        if len(self.indra_tensor) == 1:
+        if self.ndim == 1:
             return self.indra_tensor.bytes().decode()
         return list(
             self.indra_tensor[i].bytes().decode() for i in range(len(self.indra_tensor))
         )
 
     def dict(self, fetch_chunks: bool = False):
         """Return json data. Only applicable for tensors with 'json' base htype."""
-        if len(self.indra_tensor) == 1:
+        if self.ndim == 1:
             return json.loads(self.indra_tensor.bytes().decode())
         return list(
             json.loads(self.indra_tensor[i].bytes().decode())
@@ -127,6 +127,17 @@ def min_shape(self):
 
     @property
     def shape(self):
+        result = ()
+        index_0 = self.deeplake_tensor.index.values[0]
+        if self.indra_tensor.min_shape == self.indra_tensor.max_shape:
+            if index_0.is_trivial():
+                return self.indra_tensor.shape_interval
+            else:
+                return self.indra_tensor.shape_interval[1:]
+
+        if index_0.subscriptable():
+            return (len(self.indra_tensor), *self.indra_tensor.shape)  # type: ignore
+
         return self.indra_tensor.shape
 
     @property
@@ -139,7 +150,14 @@ def shape_interval(self):
 
     @property
     def ndim(self):
-        return len(self.max_shape)
+        ndim = len(self.indra_tensor.min_shape) + 1
+        if self.indra_tensor.is_sequence:
+            ndim += 1
+        if self.deeplake_tensor.index:
+            for idx in self.deeplake_tensor.index.values:
+                if not idx.subscriptable():
+                    ndim -= 1
+        return ndim
 
     @property
     def meta(self):