Fix backward compatibility issue of pad_missing in LabelListTensorizer

pytext/data/tensorizers.py

@@ -816,20 +816,20 @@ class LabelListTensorizer(LabelTensorizer):
 
     class Config(LabelTensorizer.Config):
         # pad missing label in the list, including None and empty
-        pad_missing: bool = True
+        pad_missing: bool = False
 
     @classmethod
     def from_config(cls, config: Config):
         return cls(
-            config.pad_missing,
             config.column,
             config.allow_unknown,
             config.pad_in_vocab,
             config.label_vocab,
             config.is_input,
+            pad_missing=config.pad_missing,
         )
 
-    def __init__(self, pad_missing: bool = True, *args, **kwargs):
+    def __init__(self, *args, pad_missing: bool = False, **kwargs):
         super().__init__(*args, **kwargs)
         self.pad_missing = pad_missing
 

pytext/data/test/tensorizers_test.py

@@ -144,7 +144,7 @@ def test_label_list_tensors_pad_missing(self):
         )
         tensorizers = {
             "label": LabelListTensorizer(
-                label_column="label", pad_in_vocab=False, allow_unknown=False
+                pad_missing=True, label_column="label", pad_in_vocab=False, allow_unknown=False
             )
         }
         initialize_tensorizers(tensorizers, ds.test)

pytext/data/utils.py

@@ -250,7 +250,7 @@ def add_all(self, values) -> None:
                 self.add_all(value)
         else:
             # Don't add None or empty
-            if values:
+            if values not in [None, ""]:
                 self.add(values)
 
     def add(self, value) -> None: