* Fix empty values in attributes and parts of speech, so symbols align correctly with the StringStore

spacy/parts_of_speech.pyx

@@ -2,7 +2,7 @@ from __future__ import unicode_literals
 
 
 IDS = {
-    "NO_TAG": NO_TAG,
+    "": NO_TAG,
     "ADJ": ADJ,
     "ADP": ADP,
     "ADV": ADV,

spacy/symbols.pyx

@@ -1,5 +1,4 @@
 SYMBOL_IDS = {
-    "EMPTY_VALUE": EMPTY_VALUE,
     "Attr_is_alpha": Attr_is_alpha,
     "Attr_is_ascii": Attr_is_ascii,
     "Attr_is_digit": Attr_is_digit,

spacy/vocab.pyx

@@ -76,9 +76,11 @@ cdef class Vocab:
         # strings are loaded first, because the vocab is open-class, and these
         # symbols are closed class.
         for name in attrs.NAMES:
-            _ = self.strings[name]
+            if name:
+                _ = self.strings[name]
         for name in parts_of_speech.NAMES:
-            _ = self.strings[name]
+            if name:
+                _ = self.strings[name]
         #for morph_name in UNIV_MORPH_NAMES:
         #    _ = self.strings[morph_name]
         #for entity_type_name in entity_types.NAMES:

tests/vocab/test_vocab.py

@@ -1,6 +1,9 @@
 from __future__ import unicode_literals
 import pytest
 
+from spacy.attrs import LEMMA, ORTH, PROB, IS_ALPHA
+from spacy.parts_of_speech import NOUN, VERB
+
 
 def test_neq(en_vocab):
     addr = en_vocab['Hello']
@@ -25,3 +28,13 @@ def test_punct_neq(en_vocab):
 def test_shape_attr(en_vocab):
     example = en_vocab['example']
     assert example.orth != example.shape
+
+
+def test_symbols(en_vocab):
+    assert en_vocab.strings['IS_ALPHA'] == IS_ALPHA
+    assert en_vocab.strings['NOUN'] == NOUN
+    assert en_vocab.strings['VERB'] == VERB
+    assert en_vocab.strings['LEMMA'] == LEMMA
+    assert en_vocab.strings['ORTH'] == ORTH
+    assert en_vocab.strings['PROB'] == PROB
+