fix(sqlparser): fix subclause order when parsing function calls (#18130)

Signed-off-by: Richard Chien <[email protected]>

src/frontend/planner_test/tests/testdata/output/agg.yaml

@@ -769,7 +769,7 @@
     create table t(a int, b int);
     select avg(a) FILTER (WHERE abs(a)) AS avga from t;
   binder_error: |
-    Failed to bind expression: avg(a) FILTER(WHERE abs(a))
+    Failed to bind expression: avg(a) FILTER (WHERE abs(a))
 
     Caused by:
       argument of FILTER must be boolean, not type Int32
@@ -778,7 +778,7 @@
     create table t(a int, b int);
     select avg(a) FILTER (WHERE 0 < (select max(a) from t)) AS avga from t;
   binder_error: |
-    Failed to bind expression: avg(a) FILTER(WHERE 0 < (SELECT max(a) FROM t))
+    Failed to bind expression: avg(a) FILTER (WHERE 0 < (SELECT max(a) FROM t))
 
     Caused by:
       Feature is not yet implemented: subquery in filter clause
@@ -788,7 +788,7 @@
     create table t(a int, b int);
     select avg(a) FILTER (WHERE a < avg(b)) AS avga from t;
   binder_error: |
-    Failed to bind expression: avg(a) FILTER(WHERE a < avg(b))
+    Failed to bind expression: avg(a) FILTER (WHERE a < avg(b))
 
     Caused by:
       Feature is not yet implemented: aggregation function in filter clause
@@ -798,7 +798,7 @@
     create table t(a int, b int);
     select abs(a) FILTER (WHERE a > 0) AS avga from t;
   binder_error: |
-    Failed to bind expression: abs(a) FILTER(WHERE a > 0)
+    Failed to bind expression: abs(a) FILTER (WHERE a > 0)
 
     Caused by:
       Invalid input syntax: DISTINCT, ORDER BY or FILTER is only allowed in aggregation functions, but `abs` is not an aggregation function
@@ -1096,7 +1096,7 @@
     create table a (a1 int, a2 int);
     select count(a1) filter (where (select true)) from a;
   binder_error: |
-    Failed to bind expression: count(a1) FILTER(WHERE (SELECT true))
+    Failed to bind expression: count(a1) FILTER (WHERE (SELECT true))
 
     Caused by:
       Feature is not yet implemented: subquery in filter clause
@@ -1110,10 +1110,10 @@
       select count(b1) filter (where min(a1) < 3) from b
     );
   binder_error: |
-    Failed to bind expression: EXISTS (SELECT count(b1) FILTER(WHERE min(a1) < 3) FROM b)
+    Failed to bind expression: EXISTS (SELECT count(b1) FILTER (WHERE min(a1) < 3) FROM b)
 
     Caused by these errors (recent errors listed first):
-      1: Failed to bind expression: count(b1) FILTER(WHERE min(a1) < 3)
+      1: Failed to bind expression: count(b1) FILTER (WHERE min(a1) < 3)
       2: Feature is not yet implemented: aggregation function in filter clause
     No tracking issue yet. Feel free to submit a feature request at https://github.com/risingwavelabs/risingwave/issues/new?labels=type%2Ffeature&template=feature_request.yml
 - name: agg filter - table function
@@ -1122,7 +1122,7 @@
     create table a (a1 int, a2 int);
     select count(a1) filter (where unnest(array[1]) < 1) from a;
   binder_error: |
-    Failed to bind expression: count(a1) FILTER(WHERE unnest(ARRAY[1]) < 1)
+    Failed to bind expression: count(a1) FILTER (WHERE unnest(ARRAY[1]) < 1)
 
     Caused by:
       Invalid input syntax: table functions are not allowed in FILTER
@@ -1551,7 +1551,7 @@
     create table t (x int, y int);
     select percentile_cont(x) within group (order by y) from t;
   binder_error: |
-    Failed to bind expression: percentile_cont(x)
+    Failed to bind expression: percentile_cont(x) WITHIN GROUP (ORDER BY y)
 
     Caused by:
       Feature is not yet implemented: variable as direct argument of ordered-set aggregate
@@ -1560,7 +1560,7 @@
     create table t (x int, y int);
     select percentile_cont('abc') within group (order by y) from t;
   binder_error: |
-    Failed to bind expression: percentile_cont('abc')
+    Failed to bind expression: percentile_cont('abc') WITHIN GROUP (ORDER BY y)
 
     Caused by:
       Feature is not yet implemented: variable as direct argument of ordered-set aggregate
@@ -1569,23 +1569,23 @@
     create table t (x int, y int);
     select percentile_cont(1.3) within group (order by y) from t;
   binder_error: |
-    Failed to bind expression: percentile_cont(1.3)
+    Failed to bind expression: percentile_cont(1.3) WITHIN GROUP (ORDER BY y)
 
     Caused by:
       Invalid input syntax: direct arg in `percentile_cont` must between 0.0 and 1.0
 - sql: |
     create table t (x int, y int);
     select percentile_cont(0, 0) within group (order by y) from t;
   binder_error: |
-    Failed to bind expression: percentile_cont(0, 0)
+    Failed to bind expression: percentile_cont(0, 0) WITHIN GROUP (ORDER BY y)
 
     Caused by:
       Invalid input syntax: invalid direct args or within group argument for `percentile_cont` aggregation
 - sql: |
     create table t (x int, y varchar);
     select percentile_cont(0) within group (order by y) from t;
   binder_error: |
-    Failed to bind expression: percentile_cont(0)
+    Failed to bind expression: percentile_cont(0) WITHIN GROUP (ORDER BY y)
 
     Caused by:
       Invalid input syntax: arg in `percentile_cont` must be castable to float64
@@ -1625,15 +1625,15 @@
     create table t (x int, y int);
     select percentile_cont(0.8 + 0.8) within group (order by y) from t;
   binder_error: |
-    Failed to bind expression: percentile_cont(0.8 + 0.8)
+    Failed to bind expression: percentile_cont(0.8 + 0.8) WITHIN GROUP (ORDER BY y)
 
     Caused by:
       Invalid input syntax: direct arg in `percentile_cont` must between 0.0 and 1.0
 - sql: |
     create table t (x int, y int);
     select percentile_cont(0.2 + x) within group (order by y) from t group by x;
   binder_error: |
-    Failed to bind expression: percentile_cont(0.2 + x)
+    Failed to bind expression: percentile_cont(0.2 + x) WITHIN GROUP (ORDER BY y)
 
     Caused by:
       Feature is not yet implemented: variable as direct argument of ordered-set aggregate
@@ -1656,7 +1656,7 @@
     create table t (x int, y varchar);
     select mode(1) within group (order by y desc) from t;
   binder_error: |
-    Failed to bind expression: mode(1)
+    Failed to bind expression: mode(1) WITHIN GROUP (ORDER BY y DESC)
 
     Caused by:
       Invalid input syntax: invalid direct args or within group argument for `mode` aggregation

src/frontend/planner_test/tests/testdata/output/expr.yaml

@@ -558,7 +558,7 @@
     create table t (v1 timestamp with time zone, v2 int);
     select sum(v2) filter (where v1 >= now()) as sum_v2 from t;
   stream_error: |
-    Failed to bind expression: sum(v2) FILTER(WHERE v1 >= now())
+    Failed to bind expression: sum(v2) FILTER (WHERE v1 >= now())
 
     Caused by:
       Invalid input syntax: For streaming queries, `NOW()` function is only allowed in `WHERE`, `HAVING`, `ON` and `FROM`. Found in clause: Some(Filter). Please please refer to https://www.risingwave.dev/docs/current/sql-pattern-temporal-filters/ for more information

src/frontend/planner_test/tests/testdata/output/over_window_function.yaml

@@ -50,7 +50,7 @@
     create table t(x int);
     select sum(x) filter (where row_number() over () > 1) from t;
   binder_error: |
-    Failed to bind expression: sum(x) FILTER(WHERE row_number() OVER () > 1)
+    Failed to bind expression: sum(x) FILTER (WHERE row_number() OVER () > 1)
 
     Caused by:
       Invalid input syntax: window functions are not allowed in FILTER

src/frontend/planner_test/tests/testdata/output/project_set.yaml

@@ -228,7 +228,7 @@
   sql: |
     with a(a1) as (values (array[2]), (null)) select count(*) filter(where unnest(a1) < 3) from a;
   binder_error: |
-    Failed to bind expression: count(*) FILTER(WHERE unnest(a1) < 3)
+    Failed to bind expression: count(*) FILTER (WHERE unnest(a1) < 3)
 
     Caused by:
       Invalid input syntax: table functions are not allowed in FILTER

src/frontend/src/binder/mod.rs

@@ -816,8 +816,6 @@ mod tests {
                                                 order_by: [],
                                                 ignore_nulls: false,
                                             },
-                                            over: None,
-                                            filter: None,
                                             within_group: Some(
                                                 OrderByExpr {
                                                     expr: Identifier(
@@ -830,6 +828,8 @@ mod tests {
                                                     nulls_first: None,
                                                 },
                                             ),
+                                            filter: None,
+                                            over: None,
                                         },
                                     ),
                                 ),

src/sqlparser/src/ast/mod.rs

@@ -2567,13 +2567,19 @@ impl FunctionArgList {
 #[derive(Debug, Clone, PartialEq, Eq, Hash)]
 #[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
 pub struct Function {
-    /// Whether the function is prefixed with `aggregate:`
+    /// Whether the function is prefixed with `AGGREGATE:`
     pub scalar_as_agg: bool,
+    /// Function name.
     pub name: ObjectName,
+    /// Argument list of the function call, i.e. things in `()`.
     pub arg_list: FunctionArgList,
-    pub over: Option<WindowSpec>,
-    pub filter: Option<Box<Expr>>,
+    /// `WITHIN GROUP` clause of the function call, for ordered-set aggregate functions.
+    /// FIXME(rc): why we only support one expression here?
     pub within_group: Option<Box<OrderByExpr>>,
+    /// `FILTER` clause of the function call, for aggregate and window (not supported yet) functions.
+    pub filter: Option<Box<Expr>>,
+    /// `OVER` clause of the function call, for window functions.
+    pub over: Option<WindowSpec>,
 }
 
 impl Function {
@@ -2582,9 +2588,9 @@ impl Function {
             scalar_as_agg: false,
             name,
             arg_list: FunctionArgList::empty(),
-            over: None,
-            filter: None,
             within_group: None,
+            filter: None,
+            over: None,
         }
     }
 }
@@ -2595,11 +2601,14 @@ impl fmt::Display for Function {
             write!(f, "AGGREGATE:")?;
         }
         write!(f, "{}{}", self.name, self.arg_list)?;
-        if let Some(o) = &self.over {
-            write!(f, " OVER ({})", o)?;
+        if let Some(within_group) = &self.within_group {
+            write!(f, " WITHIN GROUP (ORDER BY {})", within_group)?;
         }
         if let Some(filter) = &self.filter {
-            write!(f, " FILTER(WHERE {})", filter)?;
+            write!(f, " FILTER (WHERE {})", filter)?;
+        }
+        if let Some(o) = &self.over {
+            write!(f, " OVER ({})", o)?;
         }
         Ok(())
     }

src/sqlparser/src/parser.rs

@@ -824,6 +824,27 @@ impl Parser<'_> {
         };
         let name = self.parse_object_name()?;
         let arg_list = self.parse_argument_list()?;
+
+        let within_group = if self.parse_keywords(&[Keyword::WITHIN, Keyword::GROUP]) {
+            self.expect_token(&Token::LParen)?;
+            self.expect_keywords(&[Keyword::ORDER, Keyword::BY])?;
+            let order_by = self.parse_order_by_expr()?;
+            self.expect_token(&Token::RParen)?;
+            Some(Box::new(order_by.clone()))
+        } else {
+            None
+        };
+
+        let filter = if self.parse_keyword(Keyword::FILTER) {
+            self.expect_token(&Token::LParen)?;
+            self.expect_keyword(Keyword::WHERE)?;
+            let filter_expr = self.parse_expr()?;
+            self.expect_token(&Token::RParen)?;
+            Some(Box::new(filter_expr))
+        } else {
+            None
+        };
+
         let over = if self.parse_keyword(Keyword::OVER) {
             // TODO: support window names (`OVER mywin`) in place of inline specification
             self.expect_token(&Token::LParen)?;
@@ -855,33 +876,13 @@ impl Parser<'_> {
             None
         };
 
-        let filter = if self.parse_keyword(Keyword::FILTER) {
-            self.expect_token(&Token::LParen)?;
-            self.expect_keyword(Keyword::WHERE)?;
-            let filter_expr = self.parse_expr()?;
-            self.expect_token(&Token::RParen)?;
-            Some(Box::new(filter_expr))
-        } else {
-            None
-        };
-
-        let within_group = if self.parse_keywords(&[Keyword::WITHIN, Keyword::GROUP]) {
-            self.expect_token(&Token::LParen)?;
-            self.expect_keywords(&[Keyword::ORDER, Keyword::BY])?;
-            let order_by = self.parse_order_by_expr()?;
-            self.expect_token(&Token::RParen)?;
-            Some(Box::new(order_by.clone()))
-        } else {
-            None
-        };
-
         Ok(Expr::Function(Function {
             scalar_as_agg,
             name,
             arg_list,
-            over,
-            filter,
             within_group,
+            filter,
+            over,
         }))
     }