Fix tokenizer bug when just quotes before colon

This fixes a bug where ChemWordTokenizer raised an IndexError when it encountered a quote, followed by a colon, followed by a digit.

chemdataextractor/nlp/tokenize.py

@@ -657,7 +657,7 @@ def _subspan(self, s, span, nextspan):
             after = text[i+1:]
             if char in {':', ';'}:
                 # Split around colon unless it looks like we're in a chemical name
-                if not (before and after and after[0].isdigit() and before.rstrip('′\'')[-1].isdigit() and '-' in after) and not (self.NO_SPLIT_CHEM.search(before) and self.NO_SPLIT_CHEM.search(after)):
+                if not (before and after and after[0].isdigit() and before.rstrip('′\'')[-1:].isdigit() and '-' in after) and not (self.NO_SPLIT_CHEM.search(before) and self.NO_SPLIT_CHEM.search(after)):
                     return self._split_span(span, i, 1)
             elif char in {'x', '+', '−'}:
                 # Split around x, +, − (\u2212 minus) between two numbers or at start followed by numbers

tests/test_nlp_tokenize.py

@@ -778,6 +778,10 @@ def test_bracket_whitespace_error(self):
         self.assertEqual(['7.95', '(', 's', ',', '4H', ')'], self.t.tokenize('7.95(s, 4H)'))
         self.assertEqual(['In', 'Fig.', '5', '(', 'a', ',', 'b', ')'], self.t.tokenize('In Fig. 5(a, b)'))
 
+    def test_quote_colon(self):
+        """Test the word tokenizer quote followed by colon followed by digit (IndexError bugfix)."""
+        self.assertEqual(['\'', ':', '1'], self.t.tokenize('\':1'))
+
     def test_chemtext_sentence(self):
         """Test tokenization through the Text and Sentence API."""
         t = Text('Hi, my name is Matt. What is your name?')