Feat(postgres): add support for XMLTABLE

Author: georgesittas

sqlglot/dialects/oracle.py

@@ -144,7 +144,6 @@ class Parser(parser.Parser):
                 this=self._parse_format_json(self._parse_bitwise()),
                 order=self._parse_order(),
             ),
-            "XMLTABLE": lambda self: self._parse_xml_table(),
             "JSON_EXISTS": lambda self: self._parse_json_exists(),
         }
 
@@ -179,26 +178,6 @@ class Parser(parser.Parser):
             ),
         }
 
-        def _parse_xml_table(self) -> exp.XMLTable:
-            this = self._parse_string()
-
-            passing = None
-            columns = None
-
-            if self._match_text_seq("PASSING"):
-                # The BY VALUE keywords are optional and are provided for semantic clarity
-                self._match_text_seq("BY", "VALUE")
-                passing = self._parse_csv(self._parse_column)
-
-            by_ref = self._match_text_seq("RETURNING", "SEQUENCE", "BY", "REF")
-
-            if self._match_text_seq("COLUMNS"):
-                columns = self._parse_csv(self._parse_field_def)
-
-            return self.expression(
-                exp.XMLTable, this=this, passing=passing, columns=columns, by_ref=by_ref
-            )
-
         def _parse_json_array(self, expr_type: t.Type[E], **kwargs) -> E:
             return self.expression(
                 expr_type,
@@ -353,17 +332,6 @@ def currenttimestamp_sql(self, expression: exp.CurrentTimestamp) -> str:
         def offset_sql(self, expression: exp.Offset) -> str:
             return f"{super().offset_sql(expression)} ROWS"
 
-        def xmltable_sql(self, expression: exp.XMLTable) -> str:
-            this = self.sql(expression, "this")
-            passing = self.expressions(expression, key="passing")
-            passing = f"{self.sep()}PASSING{self.seg(passing)}" if passing else ""
-            columns = self.expressions(expression, key="columns")
-            columns = f"{self.sep()}COLUMNS{self.seg(columns)}" if columns else ""
-            by_ref = (
-                f"{self.sep()}RETURNING SEQUENCE BY REF" if expression.args.get("by_ref") else ""
-            )
-            return f"XMLTABLE({self.sep('')}{self.indent(this + passing + by_ref + columns)}{self.seg(')', sep='')}"
-
         def add_column_sql(self, expression: exp.Alter) -> str:
             actions = self.expressions(expression, key="actions", flat=True)
             if len(expression.args.get("actions", [])) > 1:

sqlglot/generator.py

@@ -4646,3 +4646,12 @@ def unpivotcolumns_sql(self, expression: exp.UnpivotColumns) -> str:
         values = self.expressions(expression, flat=True)
 
         return f"NAME {name} VALUE {values}"
+
+    def xmltable_sql(self, expression: exp.XMLTable) -> str:
+        this = self.sql(expression, "this")
+        passing = self.expressions(expression, key="passing")
+        passing = f"{self.sep()}PASSING{self.seg(passing)}" if passing else ""
+        columns = self.expressions(expression, key="columns")
+        columns = f"{self.sep()}COLUMNS{self.seg(columns)}" if columns else ""
+        by_ref = f"{self.sep()}RETURNING SEQUENCE BY REF" if expression.args.get("by_ref") else ""
+        return f"XMLTABLE({self.sep('')}{self.indent(this + passing + by_ref + columns)}{self.seg(')', sep='')}"

sqlglot/parser.py

@@ -1150,6 +1150,7 @@ class Parser(metaclass=_Parser):
             this=self._match_text_seq("NAME") and self._parse_id_var(),
             expressions=self._match(TokenType.COMMA) and self._parse_csv(self._parse_expression),
         ),
+        "XMLTABLE": lambda self: self._parse_xml_table(),
     }
 
     QUERY_MODIFIER_PARSERS = {
@@ -6152,6 +6153,26 @@ def _parse_convert(
 
         return self.expression(exp.Cast if strict else exp.TryCast, this=this, to=to, safe=safe)
 
+    def _parse_xml_table(self) -> exp.XMLTable:
+        this = self._parse_string()
+
+        passing = None
+        columns = None
+
+        if self._match_text_seq("PASSING"):
+            # The BY VALUE keywords are optional and are provided for semantic clarity
+            self._match_text_seq("BY", "VALUE")
+            passing = self._parse_csv(self._parse_column)
+
+        by_ref = self._match_text_seq("RETURNING", "SEQUENCE", "BY", "REF")
+
+        if self._match_text_seq("COLUMNS"):
+            columns = self._parse_csv(self._parse_field_def)
+
+        return self.expression(
+            exp.XMLTable, this=this, passing=passing, columns=columns, by_ref=by_ref
+        )
+
     def _parse_decode(self) -> t.Optional[exp.Decode | exp.Case]:
         """
         There are generally two variants of the DECODE function:

tests/dialects/test_postgres.py

@@ -71,6 +71,9 @@ def test_postgres(self):
         self.validate_identity("EXEC AS myfunc @id = 123", check_command_warning=True)
         self.validate_identity("SELECT CURRENT_USER")
         self.validate_identity("SELECT * FROM ONLY t1")
+        self.validate_identity(
+            "SELECT id, name FROM XMLTABLE('/root/user' PASSING xml_data COLUMNS id INT PATH '@id', name TEXT PATH 'name/text()') AS t"
+        )
         self.validate_identity(
             "SELECT * FROM t WHERE some_column >= CURRENT_DATE + INTERVAL '1 day 1 hour' AND some_another_column IS TRUE"
         )