DynamoDB: allow leading underscore in expression attr name (#6082)

moto/dynamodb/parsing/tokens.py

@@ -116,8 +116,9 @@ class ExpressionTokenizer(object):
         An expression attribute name must begin with a pound sign (#), and be followed by one or more alphanumeric
          characters.
         """
-        return input_string.startswith("#") and cls.is_expression_attribute(
-            input_string[1:]
+        return (
+            input_string.startswith("#")
+            and re.compile("^[a-zA-Z0-9_]+$").match(input_string[1:]) is not None
         )
 
     @classmethod
@@ -182,8 +183,9 @@ class ExpressionTokenizer(object):
 
     def _make_list(self) -> List[Token]:
         """
-        Just go through characters if a character is not a token boundary stage it for adding it as a grouped token
-        later if it is a tokenboundary process staged characters and then process the token boundary as well.
+        Just go through characters
+          if a character is not a token boundary,  stage it for adding it as a grouped token later
+          if it is a tokenboundary, process staged characters, and then process the token boundary as well.
         """
         for character in self.input_expression_str:
             if not self.is_possible_token_boundary(character):

tests/test_dynamodb/test_dynamodb_expression_tokenizer.py

@@ -231,6 +231,18 @@ def test_expression_tokenizer_single_set_action_attribute_name_leading_number():
     ]
 
 
+def test_expression_tokenizer_single_set_action_attribute_name_leading_underscore():
+    set_action = "SET attr=#_sth"
+    token_list = ExpressionTokenizer.make_list(set_action)
+    assert token_list == [
+        Token(Token.ATTRIBUTE, "SET"),
+        Token(Token.WHITESPACE, " "),
+        Token(Token.ATTRIBUTE, "attr"),
+        Token(Token.EQUAL_SIGN, "="),
+        Token(Token.ATTRIBUTE_NAME, "#_sth"),
+    ]
+
+
 def test_expression_tokenizer_just_a_pipe():
     set_action = "|"
     try: