Merge upstream main

Author: revitalkr

src/ast/mod.rs

@@ -8023,6 +8023,11 @@ pub enum FunctionArgOperator {
     Colon,
     /// function(arg1 VALUE value1)
     Value,
+    /// function(arg1 value1), with no operator between the name and the value,
+    /// as in PostgreSQL `XMLPARSE(DOCUMENT value)`
+    ///
+    /// [PostgreSQL](https://www.postgresql.org/docs/current/datatype-xml.html#DATATYPE-XML-CREATING)
+    Space,
 }
 
 impl fmt::Display for FunctionArgOperator {
@@ -8033,6 +8038,7 @@ impl fmt::Display for FunctionArgOperator {
             FunctionArgOperator::Assignment => f.write_str(":="),
             FunctionArgOperator::Colon => f.write_str(":"),
             FunctionArgOperator::Value => f.write_str("VALUE"),
+            FunctionArgOperator::Space => Ok(()),
         }
     }
 }
@@ -8075,17 +8081,31 @@ impl fmt::Display for FunctionArg {
                 name,
                 arg,
                 operator,
-            } => write!(f, "{name} {operator} {arg}"),
+            } => fmt_named_function_arg(f, name, operator, arg),
             FunctionArg::ExprNamed {
                 name,
                 arg,
                 operator,
-            } => write!(f, "{name} {operator} {arg}"),
+            } => fmt_named_function_arg(f, name, operator, arg),
             FunctionArg::Unnamed(unnamed_arg) => write!(f, "{unnamed_arg}"),
         }
     }
 }
 
+/// `FunctionArgOperator::Space` has no token of its own, so the name and the
+/// value are separated by a single space instead.
+fn fmt_named_function_arg(
+    f: &mut fmt::Formatter,
+    name: &impl fmt::Display,
+    operator: &FunctionArgOperator,
+    arg: &FunctionArgExpr,
+) -> fmt::Result {
+    match operator {
+        FunctionArgOperator::Space => write!(f, "{name} {arg}"),
+        _ => write!(f, "{name} {operator} {arg}"),
+    }
+}
+
 #[derive(Debug, Clone, PartialEq, PartialOrd, Eq, Ord, Hash)]
 #[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
 #[cfg_attr(feature = "visitor", derive(Visit, VisitMut))]

src/dialect/bigquery.rs

@@ -136,6 +136,14 @@ impl Dialect for BigQueryDialect {
         true
     }
 
+    /// BigQuery allows a query to start with `FROM` (e.g. `FROM t`, and the
+    /// entry form for pipe syntax, `FROM t |> ...`).
+    ///
+    /// See <https://cloud.google.com/bigquery/docs/reference/standard-sql/query-syntax#from_queries>
+    fn supports_from_first_select(&self) -> bool {
+        true
+    }
+
     /// See <https://cloud.google.com/bigquery/docs/reference/standard-sql/procedural-language#execute_immediate>
     fn supports_execute_immediate(&self) -> bool {
         true

src/parser/mod.rs

@@ -2531,6 +2531,23 @@ impl<'a> Parser<'a> {
         })
     }
 
+    /// Parse the argument list of `XMLPARSE({ DOCUMENT | CONTENT } value)`,
+    /// including the closing parenthesis. The mode word becomes the name of the
+    /// single argument, with no operator between it and the value.
+    fn parse_xmlparse_argument_list(&mut self) -> Result<FunctionArgumentList, ParserError> {
+        let arg = FunctionArg::Named {
+            name: self.parse_identifier()?,
+            arg: FunctionArgExpr::Expr(self.parse_expr()?),
+            operator: FunctionArgOperator::Space,
+        };
+        self.expect_token(&Token::RParen)?;
+        Ok(FunctionArgumentList {
+            duplicate_treatment: None,
+            args: vec![arg],
+            clauses: vec![],
+        })
+    }
+
     /// Parse a function call expression named by `name` and return it as an `Expr`.
     pub fn parse_function(&mut self, name: ObjectName) -> Result<Expr, ParserError> {
         self.parse_function_call(name).map(Expr::Function)
@@ -2556,7 +2573,13 @@ impl<'a> Parser<'a> {
             });
         }
 
-        let mut args = self.parse_function_argument_list()?;
+        let mut args = if self.dialect.supports_xml_expressions()
+            && Self::is_simple_unquoted_object_name(&name, "xmlparse")
+        {
+            self.parse_xmlparse_argument_list()?
+        } else {
+            self.parse_function_argument_list()?
+        };
         let mut parameters = FunctionArguments::None;
         // ClickHouse aggregations support parametric functions like `HISTOGRAM(0.5, 0.6)(x, y)`
         // which (0.5, 0.6) is a parameter to the function.

tests/sqlparser_bigquery.rs

@@ -2950,3 +2950,10 @@ fn test_create_snapshot_table() {
         "CREATE SNAPSHOT TABLE IF NOT EXISTS dataset_id.table1 CLONE dataset_id.table2 FOR SYSTEM_TIME AS OF TIMESTAMP_SUB(CURRENT_TIMESTAMP(), INTERVAL 1 HOUR) OPTIONS(expiration_timestamp = TIMESTAMP '2025-01-01 00:00:00 UTC')",
     );
 }
+
+#[test]
+fn parse_from_first_select() {
+    bigquery().verified_stmt("FROM t");
+    bigquery().verified_stmt("FROM t SELECT a, b");
+    bigquery().verified_stmt("FROM t |> WHERE a > 1 |> SELECT a");
+}

tests/sqlparser_common.rs

@@ -19604,6 +19604,56 @@ fn parse_aliased_function_args() {
         .is_err());
 }
 
+#[test]
+fn parse_xmlparse() {
+    let dialects = all_dialects_where(|d| d.supports_xml_expressions());
+
+    for (sql, mode) in [
+        ("SELECT xmlparse(content '<a/>')", "content"),
+        ("SELECT xmlparse(document '<a/>')", "document"),
+    ] {
+        let select = dialects.verified_only_select(sql);
+        match expr_from_projection(&select.projection[0]) {
+            Expr::Function(Function {
+                name,
+                args: FunctionArguments::List(list),
+                ..
+            }) => {
+                assert_eq!(name.to_string(), "xmlparse");
+                assert_eq!(
+                    list.args,
+                    vec![FunctionArg::Named {
+                        name: Ident::new(mode),
+                        arg: FunctionArgExpr::Expr(Expr::Value(
+                            Value::SingleQuotedString("<a/>".to_string()).into()
+                        )),
+                        operator: FunctionArgOperator::Space,
+                    }]
+                );
+            }
+            expr => panic!("expected an XMLPARSE function call, got {expr:?}"),
+        }
+    }
+
+    // XMLPARSE needs both a mode word and a value.
+    assert!(dialects
+        .parse_sql_statements("SELECT xmlparse('<a/>')")
+        .is_err());
+
+    // Going through the ordinary function-call path, XMLPARSE accepts the same
+    // trailing clauses as any other function.
+    dialects.verified_stmt("SELECT xmlparse(document x) FILTER (WHERE y > 1)");
+    dialects.verified_stmt("SELECT xmlparse(document x) OVER (PARTITION BY y)");
+
+    // On dialects without XML support, `xmlparse` stays a regular function
+    // and the special `CONTENT <expr>` syntax is rejected.
+    let others = all_dialects_except(|d| d.supports_xml_expressions());
+    others.verified_only_select("SELECT xmlparse(1)");
+    assert!(others
+        .parse_sql_statements("SELECT xmlparse(content '<a/>')")
+        .is_err());
+}
+
 /// Regression test for the 2^N parse-time blowup in `parse_compound_expr` on
 /// inputs like `IF a0.a1...aN.#`. The parse is run on a worker thread and the
 /// main thread asserts that it reports back within a generous timeout. Post-fix

tests/sqlparser_postgres.rs

@@ -4130,6 +4130,21 @@ fn parse_xmlforest_aliased_arguments() {
     );
 }
 
+#[test]
+fn parse_xmlparse() {
+    // The parser only distinguishes the two modes, so the corpus covers those
+    // plus a non-literal argument.
+    let statements = [
+        "SELECT XMLPARSE(CONTENT '')",
+        "SELECT XMLPARSE(CONTENT '<abc>x</abc>')",
+        "SELECT XMLPARSE(DOCUMENT '<abc>x</abc>')",
+        "SELECT XMLPARSE(DOCUMENT col || '</abc>')",
+    ];
+    for sql in statements {
+        pg().verified_stmt(sql);
+    }
+}
+
 #[test]
 fn parse_xml_typed_string() {
     // xml '...' should parse as a TypedString on PostgreSQL and Generic