Refactor: clean up logic related to join marker parsing/generation (#3613)

georgesittas · web-flow · commit 6194c0f37fd3 · 2024-06-07T20:57:55.000+03:00
diff --git a/sqlglot/dialects/dialect.py b/sqlglot/dialects/dialect.py
@@ -220,6 +220,9 @@ class Dialect(metaclass=_Dialect):
     SUPPORTS_SEMI_ANTI_JOIN = True
     """Whether `SEMI` or `ANTI` joins are supported."""
 
+    SUPPORTS_COLUMN_JOIN_MARKS = False
+    """Whether the old-style outer join (+) syntax is supported."""
+
     NORMALIZE_FUNCTIONS: bool | str = "upper"
     """
     Determines how function names are going to be normalized.
diff --git a/sqlglot/dialects/oracle.py b/sqlglot/dialects/oracle.py
@@ -36,6 +36,7 @@ class Oracle(Dialect):
     ALIAS_POST_TABLESAMPLE = True
     LOCKING_READS_SUPPORTED = True
     TABLESAMPLE_SIZE_IS_PERCENT = True
+    SUPPORTS_COLUMN_JOIN_MARKS = True
 
     # See section 8: https://docs.oracle.com/cd/A97630_01/server.920/a96540/sql_elements9a.htm
     NORMALIZATION_STRATEGY = NormalizationStrategy.UPPERCASE
@@ -173,12 +174,6 @@ def _parse_json_array(self, expr_type: t.Type[E], **kwargs) -> E:
                 **kwargs,
             )
 
-        def _parse_column(self) -> t.Optional[exp.Expression]:
-            column = super()._parse_column()
-            if column:
-                column.set("join_mark", self._match(TokenType.JOIN_MARKER))
-            return column
-
         def _parse_hint(self) -> t.Optional[exp.Hint]:
             if self._match(TokenType.HINT):
                 start = self._curr
@@ -197,7 +192,6 @@ class Generator(generator.Generator):
         LOCKING_READS_SUPPORTED = True
         JOIN_HINTS = False
         TABLE_HINTS = False
-        COLUMN_JOIN_MARKS_SUPPORTED = True
         DATA_TYPE_SPECIFIERS_ALLOWED = True
         ALTER_TABLE_INCLUDE_COLUMN_KEYWORD = False
         LIMIT_FETCH = "FETCH"
diff --git a/sqlglot/dialects/redshift.py b/sqlglot/dialects/redshift.py
@@ -40,6 +40,7 @@ class Redshift(Postgres):
     INDEX_OFFSET = 0
     COPY_PARAMS_ARE_CSV = False
     HEX_LOWERCASE = True
+    SUPPORTS_COLUMN_JOIN_MARKS = True
 
     TIME_FORMAT = "'YYYY-MM-DD HH:MI:SS'"
     TIME_MAPPING = {
@@ -115,26 +116,20 @@ def _parse_approximate_count(self) -> t.Optional[exp.ApproxDistinct]:
             self._retreat(index)
             return None
 
-        def _parse_column(self) -> t.Optional[exp.Expression]:
-            column = super()._parse_column()
-            if column:
-                column.set("join_mark", self._match(TokenType.JOIN_MARKER))
-            return column
-
     class Tokenizer(Postgres.Tokenizer):
         BIT_STRINGS = []
         HEX_STRINGS = []
         STRING_ESCAPES = ["\\", "'"]
 
         KEYWORDS = {
             **Postgres.Tokenizer.KEYWORDS,
+            "(+)": TokenType.JOIN_MARKER,
             "HLLSKETCH": TokenType.HLLSKETCH,
+            "MINUS": TokenType.EXCEPT,
             "SUPER": TokenType.SUPER,
             "TOP": TokenType.TOP,
             "UNLOAD": TokenType.COMMAND,
             "VARBYTE": TokenType.VARBINARY,
-            "MINUS": TokenType.EXCEPT,
-            "(+)": TokenType.JOIN_MARKER,
         }
         KEYWORDS.pop("VALUES")
 
@@ -155,7 +150,6 @@ class Generator(Postgres.Generator):
         HEX_FUNC = "TO_HEX"
         # Redshift doesn't have `WITH` as part of their with_properties so we remove it
         WITH_PROPERTIES_PREFIX = " "
-        COLUMN_JOIN_MARKS_SUPPORTED = True
 
         TYPE_MAPPING = {
             **Postgres.Generator.TYPE_MAPPING,
diff --git a/sqlglot/generator.py b/sqlglot/generator.py
@@ -225,9 +225,6 @@ class Generator(metaclass=_Generator):
     # Whether to generate INSERT INTO ... RETURNING or INSERT INTO RETURNING ...
     RETURNING_END = True
 
-    # Whether to generate the (+) suffix for columns used in old-style join conditions
-    COLUMN_JOIN_MARKS_SUPPORTED = False
-
     # Whether to generate an unquoted value for EXTRACT's date part argument
     EXTRACT_ALLOWS_QUOTES = True
 
@@ -827,7 +824,7 @@ def column_parts(self, expression: exp.Column) -> str:
     def column_sql(self, expression: exp.Column) -> str:
         join_mark = " (+)" if expression.args.get("join_mark") else ""
 
-        if join_mark and not self.COLUMN_JOIN_MARKS_SUPPORTED:
+        if join_mark and not self.dialect.SUPPORTS_COLUMN_JOIN_MARKS:
             join_mark = ""
             self.unsupported("Outer join syntax using the (+) operator is not supported.")
 
diff --git a/sqlglot/parser.py b/sqlglot/parser.py
@@ -4496,7 +4496,12 @@ def _parse_at_time_zone(self, this: t.Optional[exp.Expression]) -> t.Optional[ex
 
     def _parse_column(self) -> t.Optional[exp.Expression]:
         this = self._parse_column_reference()
-        return self._parse_column_ops(this) if this else self._parse_bracket(this)
+        column = self._parse_column_ops(this) if this else self._parse_bracket(this)
+
+        if self.dialect.SUPPORTS_COLUMN_JOIN_MARKS and column:
+            column.set("join_mark", self._match(TokenType.JOIN_MARKER))
+
+        return column
 
     def _parse_column_reference(self) -> t.Optional[exp.Expression]:
         this = self._parse_field()
