fix(query): allow window expr in group by query (#17962)

* chore(query): allow window expr in group by query

* chore(query): allow window expr in group by query

* chore(query): allow window expr in group by query

Author: sundy-li

.github/workflows/release.yml

@@ -417,13 +417,13 @@ jobs:
           version: ${{ needs.create_release.outputs.version }}
           arch: ${{ matrix.arch }}
 
-  # bindings_python:
-  #   if: needs.create_release.outputs.type == 'stable'
-  #   needs: create_release
-  #   uses: ./.github/workflows/bindings.python.yml
-  #   secrets: inherit
-  #   with:
-  #     tag: ${{ needs.create_release.outputs.version }}
+  bindings_python:
+    # if: needs.create_release.outputs.type == 'stable'
+    needs: create_release
+    uses: ./.github/workflows/bindings.python.yml
+    secrets: inherit
+    with:
+      tag: ${{ needs.create_release.outputs.version }}
 
   notify:
     runs-on: ubuntu-latest

src/query/functions/src/lib.rs

@@ -48,8 +48,10 @@ pub fn is_builtin_function(name: &str) -> bool {
 // The plan of search function, async function and udf contains some arguments defined in meta,
 // which may be modified by user at any time. Those functions are not not suitable for caching.
 pub fn is_cacheable_function(name: &str) -> bool {
+    let n = name;
     let name = Ascii::new(name);
-    BUILTIN_FUNCTIONS.contains(name.into_inner())
+    (BUILTIN_FUNCTIONS.contains(name.into_inner())
+        && !BUILTIN_FUNCTIONS.get_property(n).unwrap().non_deterministic)
         || AggregateFunctionFactory::instance().contains(name.into_inner())
         || GENERAL_WINDOW_FUNCTIONS.contains(&name)
         || GENERAL_LAMBDA_FUNCTIONS.contains(&name)

src/query/sql/src/planner/binder/window.rs

@@ -282,6 +282,12 @@ impl WindowInfo {
                 .insert(window.display_name.clone(), i);
         }
     }
+
+    pub fn get_window_info(&self, display_name: &str) -> Option<&WindowFunctionInfo> {
+        self.window_functions_map
+            .get(display_name)
+            .map(|index| &self.window_functions[*index])
+    }
 }
 
 #[derive(Clone, PartialEq, Eq, Debug)]

src/query/sql/src/planner/semantic/grouping_check.rs

@@ -206,6 +206,16 @@ impl VisitorMut<'_> for GroupingChecker<'_> {
             .aggregate_info
             .get_aggregate_function(&column.column.column_name)
             .is_some()
+        {
+            // Be replaced by `AggregateRewriter`.
+            return Ok(());
+        }
+
+        if self
+            .bind_context
+            .windows
+            .get_window_info(&column.column.column_name)
+            .is_some()
         {
             // Be replaced by `WindowRewriter`.
             return Ok(());

tests/sqllogictests/suites/query/aggregate.test

@@ -105,6 +105,16 @@ SELECT AVG(1 + a), sum(1 + a), count(1 + a) FROM tc;
 ----
 3.0 9 3
 
+## Aggregate with alias and window function
+statement ok
+select avg(1+a) score, a, percent_rank() over (partition by a % 3 order by score) d, d + 3 from tc group by a;
+
+statement ok
+select avg(1+a) score, a, percent_rank() over (partition by a % 3 order by a) d, d + 3 from tc group by a;
+
+statement error
+select avg(1+a) score, a, percent_rank() over (partition by a % 3 order by b) d, d + 3 from tc group by a;
+
 ## Create dimension tables with generic names
 statement ok
 CREATE or replace TABLE date_ranges (