[pull] main from apache:main

benchmarks/queries/clickbench/README.md

@@ -228,6 +228,22 @@ Results look like
 Elapsed 30.195 seconds.
 ```
 
+
+### Q9-Q12: FIRST_VALUE Aggregation Performance
+
+These queries test the performance of the `FIRST_VALUE` aggregation function with different data types and grouping cardinalities.
+
+| Query | `FIRST_VALUE` Column | Column Type | Group By Column | Group By Type | Number of Groups |
+|-------|----------------------|-------------|-----------------|---------------|------------------|
+| Q9    | `URL`                | `Utf8`      | `UserID`        | `Int64`       | 17,630,976       |
+| Q10    | `URL`                | `Utf8`      | `OS`            | `Int16`       | 91               |
+| Q11   | `WatchID`            | `Int64`     | `UserID`        | `Int64`       | 17,630,976       |
+| Q12   | `WatchID`            | `Int64`     | `OS`            | `Int16`       | 91               |
+
+
+
+
+
 ## Data Notes
 
 Here are some interesting statistics about the data used in the queries

benchmarks/queries/clickbench/extended/q10.sql

@@ -0,0 +1,8 @@
+-- Must set for ClickBench hits_partitioned dataset. See https://github.com/apache/datafusion/issues/16591
+-- set datafusion.execution.parquet.binary_as_string = true
+
+SELECT MAX(len) FROM (
+    SELECT LENGTH(FIRST_VALUE("URL" ORDER BY "EventTime")) as len
+    FROM hits
+    GROUP BY "OS"
+);

benchmarks/queries/clickbench/extended/q11.sql

@@ -0,0 +1,8 @@
+-- Must set for ClickBench hits_partitioned dataset. See https://github.com/apache/datafusion/issues/16591
+-- set datafusion.execution.parquet.binary_as_string = true
+
+SELECT MAX(fv) FROM (
+    SELECT FIRST_VALUE("WatchID" ORDER BY "EventTime") as fv
+    FROM hits
+    GROUP BY "UserID"
+);

benchmarks/queries/clickbench/extended/q12.sql

@@ -0,0 +1,8 @@
+-- Must set for ClickBench hits_partitioned dataset. See https://github.com/apache/datafusion/issues/16591
+-- set datafusion.execution.parquet.binary_as_string = true
+
+SELECT MAX(fv) FROM (
+    SELECT FIRST_VALUE("WatchID" ORDER BY "EventTime") as fv
+    FROM hits
+    GROUP BY "OS"
+);

benchmarks/queries/clickbench/extended/q8.sql

@@ -0,0 +1,4 @@
+-- Must set for ClickBench hits_partitioned dataset. See https://github.com/apache/datafusion/issues/16591
+-- set datafusion.execution.parquet.binary_as_string = true
+
+SELECT "RegionID", "UserAgent", "OS", AVG(to_timestamp("ResponseEndTiming")-to_timestamp("ResponseStartTiming")) as avg_response_time, AVG(to_timestamp("ResponseEndTiming")-to_timestamp("ConnectTiming")) as avg_latency FROM hits GROUP BY "RegionID", "UserAgent", "OS" ORDER BY avg_latency DESC limit 10;

benchmarks/queries/clickbench/extended/q9.sql

@@ -0,0 +1,8 @@
+-- Must set for ClickBench hits_partitioned dataset. See https://github.com/apache/datafusion/issues/16591
+-- set datafusion.execution.parquet.binary_as_string = true
+
+SELECT MAX(len) FROM (
+    SELECT LENGTH(FIRST_VALUE("URL" ORDER BY "EventTime")) as len
+    FROM hits
+    GROUP BY "UserID"
+);

datafusion/common/src/join_type.rs

@@ -142,6 +142,20 @@ impl JoinType {
                 | JoinType::RightMark
         )
     }
+
+    /// Returns true when an empty build side necessarily produces an empty
+    /// result for this join type.
+    pub fn empty_build_side_produces_empty_result(self) -> bool {
+        matches!(
+            self,
+            JoinType::Inner
+                | JoinType::Left
+                | JoinType::LeftSemi
+                | JoinType::LeftAnti
+                | JoinType::LeftMark
+                | JoinType::RightSemi
+        )
+    }
 }
 
 impl Display for JoinType {

datafusion/expr-common/src/type_coercion/binary.rs

@@ -1576,6 +1576,7 @@ fn string_concat_coercion(lhs_type: &DataType, rhs_type: &DataType) -> Option<Da
         (Dictionary(_, lhs_value_type), Dictionary(_, rhs_value_type)) => {
             string_coercion(lhs_value_type, rhs_value_type).or(None)
         }
+        (Binary, Binary) => Some(Utf8),
         _ => None,
     })
 }