Allowing unescaped keywords as identifiers (#82)

* do not escape some keywords

* test

* bump version

Author: StpMax

mindsdb_sql_parser/__about__.py

@@ -1,6 +1,6 @@
 __title__ = 'mindsdb_sql_parser'
 __package_name__ = 'mindsdb_sql_parser'
-__version__ = '0.13.6'
+__version__ = '0.13.7'
 __description__ = "Mindsdb SQL parser"
 __email__ = "jorge@mindsdb.com"
 __author__ = 'MindsDB Inc'

mindsdb_sql_parser/ast/select/identifier.py

@@ -21,29 +21,20 @@ def path_str_to_parts(path_str: str):
     return parts, is_quoted
 
 
-RESERVED_KEYWORDS = {
-    'PERSIST', 'IF', 'EXISTS', 'NULLS', 'FIRST', 'LAST',
-    'ORDER', 'BY', 'GROUP', 'PARTITION'
+# Here is a hardcoded set of keywords that can be used as identifiers without escaping.
+# For example, in a query like this: select {keyword} from tbl
+# If there is a need to update this list, an example code to retrieve all keywords can be found here in v0.13.6
+keywords_to_escape = {
+    "VALUES", "DESCRIBE", "THEN", "WRITE", "WITH", "INSERT", "DROP", "CROSS",
+    "SET", "ASC", "IS", "IN", "NOT", "INTO", "WINDOW", "ALTER", "WHERE",
+    "DISTINCT", "USE", "INNER", "COLLATE", "FOR", "USING", "FULL", "LIKE",
+    "JOIN", "SELECT", "OVER", "CASE", "LIMIT", "END", "UNION", "DELETE",
+    "HAVING", "OUTER", "FROM", "AS", "CHARACTER", "INTERSECT", "CONVERT",
+    "WHEN", "OR", "AND", "UPDATE", "BETWEEN", "DESC", "EXPLAIN", "SHOW",
+    "EXCEPT", "LEFT", "ELSE", "READ", "RIGHT"
 }
 
 
-_reserved_keywords: set[str] = None
-
-
-def get_reserved_words() -> set[str]:
-    global _reserved_keywords
-
-    if _reserved_keywords is None:
-        from mindsdb_sql_parser.lexer import MindsDBLexer
-
-        _reserved_keywords = RESERVED_KEYWORDS
-        for word in MindsDBLexer.tokens:
-            if '_' not in word:
-                # exclude combinations
-                _reserved_keywords.add(word)
-    return _reserved_keywords
-
-
 class Identifier(ASTNode):
     def __init__(
             self, path_str=None, parts=None, is_outer=False, with_rollup=False,
@@ -77,15 +68,14 @@ def append(self, other: "Identifier") -> None:
         self.is_quoted += other.is_quoted
 
     def iter_parts_str(self):
-        reserved_words = get_reserved_words()
         for part, is_quoted in zip(self.parts, self.is_quoted):
             if isinstance(part, Star):
                 part = str(part)
             else:
                 if (
                     is_quoted
                     or not no_wrap_identifier_regex.fullmatch(part)
-                    or part.upper() in reserved_words
+                    or part.upper() in keywords_to_escape
                 ):
                     part = f'`{part}`'
             yield part

tests/test_base_sql/test_select_structure.py

@@ -738,7 +738,29 @@ def test_partial_backticks(self):
         sql = "SELECT `integration`.`some table`.column"
         ast = parse_sql(sql)
 
-        expected_ast = Select(targets=[Identifier(parts=['integration', 'some table', 'column']),],)
+        expected_ast = Select(
+            targets=[
+                Identifier(
+                    parts=['integration', 'some table', 'column'],
+                    is_quoted=[True, True, False]
+                ),
+            ],
+        )
+
+        assert ast.to_tree() == expected_ast.to_tree()
+        assert str(ast) == str(expected_ast)
+
+    def test_keyword_escaping(self):
+        sql = "select ID, `ID`, `VALUES`"
+        ast = parse_sql(sql)
+
+        expected_ast = Select(
+            targets=[
+                Identifier(parts=['ID'], is_quoted=[False]),
+                Identifier(parts=['ID'], is_quoted=[True]),
+                Identifier(parts=['VALUES'], is_quoted=[True]),
+            ],
+        )
 
         assert ast.to_tree() == expected_ast.to_tree()
         assert str(ast) == str(expected_ast)