Fix performance issue in v2 tolist. (#1418)

src/awkward/_v2/contents/bitmaskedarray.py

@@ -634,11 +634,12 @@ def _to_list(self, behavior):
             return out
 
         mask = self.mask_as_bool(valid_when=True, nplike=self.nplike)[: self._length]
-        content = self._content._to_list(behavior)
-        out = [None] * self._length
+        out = self._content._getitem_range(slice(0, self._length))._to_list(behavior)
+
         for i, isvalid in enumerate(mask):
-            if isvalid:
-                out[i] = content[i]
+            if not isvalid:
+                out[i] = None
+
         return out
 
     def _to_nplike(self, nplike):

src/awkward/_v2/contents/bytemaskedarray.py

@@ -1035,11 +1035,12 @@ def _to_list(self, behavior):
             return out
 
         mask = self.mask_as_bool(valid_when=True, nplike=self.nplike)
-        content = self._content._to_list(behavior)
-        out = [None] * self._mask.length
+        out = self._content._getitem_range(slice(0, len(mask)))._to_list(behavior)
+
         for i, isvalid in enumerate(mask):
-            if isvalid:
-                out[i] = content[i]
+            if not isvalid:
+                out[i] = None
+
         return out
 
     def _to_nplike(self, nplike):

src/awkward/_v2/contents/indexedarray.py

@@ -1221,11 +1221,8 @@ def _to_list(self, behavior):
             return out
 
         index = self._index.raw(numpy)
-        content = self._content._to_list(behavior)
-        out = [None] * index.length
-        for i, ind in enumerate(index):
-            out[i] = content[ind]
-        return out
+        nextcontent = self._content._carry(ak._v2.index.Index(index), False)
+        return nextcontent._to_list(behavior)
 
     def _to_nplike(self, nplike):
         index = self._index._to_nplike(nplike)

src/awkward/_v2/contents/indexedoptionarray.py

@@ -1653,11 +1653,17 @@ def _to_list(self, behavior):
             return out
 
         index = self._index.raw(numpy)
-        content = self._content._to_list(behavior)
-        out = [None] * len(index)
-        for i, ind in enumerate(index):
-            if ind >= 0:
-                out[i] = content[ind]
+        not_missing = index >= 0
+
+        nextcontent = self._content._carry(
+            ak._v2.index.Index(index[not_missing]), False
+        )
+        out = nextcontent._to_list(behavior)
+
+        for i, isvalid in enumerate(not_missing):
+            if not isvalid:
+                out.insert(i, None)
+
         return out
 
     def _to_nplike(self, nplike):

src/awkward/_v2/contents/listoffsetarray.py

@@ -2062,33 +2062,48 @@ def packed(self):
         )
 
     def _to_list(self, behavior):
+        starts, stops = self.starts, self.stops
+        starts_data = starts.raw(numpy)
+        stops_data = stops.raw(numpy)[: len(starts_data)]
+
+        nonempty = starts_data != stops_data
+        if numpy.count_nonzero(nonempty) == 0:
+            mini, maxi = 0, 0
+        else:
+            mini = starts_data.min()
+            maxi = stops_data.max()
+
+        starts_data = starts_data - mini
+        stops_data = stops_data - mini
+
+        nextcontent = self._content._getitem_range(slice(mini, maxi))
+
         if self.parameter("__array__") == "bytestring":
-            content = ak._v2._util.tobytes(self._content.data)
-            starts, stops = self.starts, self.stops
+            content = ak._v2._util.tobytes(nextcontent.data)
             out = [None] * starts.length
             for i in range(starts.length):
-                out[i] = content[starts[i] : stops[i]]
+                out[i] = content[starts_data[i] : stops_data[i]]
             return out
 
         elif self.parameter("__array__") == "string":
-            content = ak._v2._util.tobytes(self._content.data)
-            starts, stops = self.starts, self.stops
+            content = ak._v2._util.tobytes(nextcontent.data)
             out = [None] * starts.length
             for i in range(starts.length):
-                out[i] = content[starts[i] : stops[i]].decode(errors="surrogateescape")
+                out[i] = content[starts_data[i] : stops_data[i]].decode(
+                    errors="surrogateescape"
+                )
             return out
 
         else:
             out = self._to_list_custom(behavior)
             if out is not None:
                 return out
 
-            content = self._content._to_list(behavior)
-            starts, stops = self.starts, self.stops
+            content = nextcontent._to_list(behavior)
             out = [None] * starts.length
 
             for i in range(starts.length):
-                out[i] = content[starts[i] : stops[i]]
+                out[i] = content[starts_data[i] : stops_data[i]]
             return out
 
     def _to_nplike(self, nplike):

src/awkward/_v2/highlevel.py

@@ -437,6 +437,13 @@ def fields(self):
         """
         return self._layout.fields
 
+    @property
+    def is_tuple(self):
+        """
+        If True, the top-most record structure has no named fields, i.e. it's a tuple.
+        """
+        return self._layout.is_tuple
+
     def _ipython_key_completions_(self):
         return self._layout.fields
 
@@ -1580,6 +1587,13 @@ def fields(self):
         """
         return self._layout.array.fields
 
+    @property
+    def is_tuple(self):
+        """
+        If True, the top-most record structure has no named fields, i.e. it's a tuple.
+        """
+        return self._layout.array.is_tuple
+
     def _ipython_key_completions_(self):
         return self._layout.array.fields