Merging upstream version 10.5.10.

Signed-off-by: Daniel Baumann <daniel@debian.org>

sqlglot/tokens.py

@@ -84,6 +84,8 @@ class TokenType(AutoName):
     TEXT = auto()
     MEDIUMTEXT = auto()
     LONGTEXT = auto()
+    MEDIUMBLOB = auto()
+    LONGBLOB = auto()
     BINARY = auto()
     VARBINARY = auto()
     JSON = auto()
@@ -587,6 +589,7 @@ class Tokenizer(metaclass=_Tokenizer):
         "PRECEDING": TokenType.PRECEDING,
         "PRIMARY KEY": TokenType.PRIMARY_KEY,
         "PROCEDURE": TokenType.PROCEDURE,
+        "QUALIFY": TokenType.QUALIFY,
         "RANGE": TokenType.RANGE,
         "RECURSIVE": TokenType.RECURSIVE,
         "REGEXP": TokenType.RLIKE,
@@ -726,6 +729,8 @@ class Tokenizer(metaclass=_Tokenizer):
         TokenType.SHOW,
     }
 
+    COMMAND_PREFIX_TOKENS = {TokenType.SEMICOLON, TokenType.BEGIN}
+
     # handle numeric literals like in hive (3L = BIGINT)
     NUMERIC_LITERALS: t.Dict[str, str] = {}
     ENCODE: t.Optional[str] = None
@@ -842,8 +847,10 @@ class Tokenizer(metaclass=_Tokenizer):
         )
         self._comments = []
 
+        # If we have either a semicolon or a begin token before the command's token, we'll parse
+        # whatever follows the command's token as a string
         if token_type in self.COMMANDS and (
-            len(self.tokens) == 1 or self.tokens[-2].token_type == TokenType.SEMICOLON
+            len(self.tokens) == 1 or self.tokens[-2].token_type in self.COMMAND_PREFIX_TOKENS
         ):
             start = self._current
             tokens = len(self.tokens)