Support Duration in min/max agg functions (#15310)

* Support min/max agg functions on values of duration type

* Add tests for GROUP BY as well

---------

Co-authored-by: svranesevic <[email protected]>
Co-authored-by: Andrew Lamb <[email protected]>

Author: svranesevic

datafusion/functions-aggregate/src/min_max.rs

@@ -22,8 +22,9 @@ mod min_max_bytes;
 
 use arrow::array::{
     ArrayRef, BinaryArray, BinaryViewArray, BooleanArray, Date32Array, Date64Array,
-    Decimal128Array, Decimal256Array, Float16Array, Float32Array, Float64Array,
-    Int16Array, Int32Array, Int64Array, Int8Array, IntervalDayTimeArray,
+    Decimal128Array, Decimal256Array, DurationMicrosecondArray, DurationMillisecondArray,
+    DurationNanosecondArray, DurationSecondArray, Float16Array, Float32Array,
+    Float64Array, Int16Array, Int32Array, Int64Array, Int8Array, IntervalDayTimeArray,
     IntervalMonthDayNanoArray, IntervalYearMonthArray, LargeBinaryArray,
     LargeStringArray, StringArray, StringViewArray, Time32MillisecondArray,
     Time32SecondArray, Time64MicrosecondArray, Time64NanosecondArray,
@@ -518,6 +519,33 @@ macro_rules! min_max_batch {
                     $OP
                 )
             }
+            DataType::Duration(TimeUnit::Second) => {
+                typed_min_max_batch!($VALUES, DurationSecondArray, DurationSecond, $OP)
+            }
+            DataType::Duration(TimeUnit::Millisecond) => {
+                typed_min_max_batch!(
+                    $VALUES,
+                    DurationMillisecondArray,
+                    DurationMillisecond,
+                    $OP
+                )
+            }
+            DataType::Duration(TimeUnit::Microsecond) => {
+                typed_min_max_batch!(
+                    $VALUES,
+                    DurationMicrosecondArray,
+                    DurationMicrosecond,
+                    $OP
+                )
+            }
+            DataType::Duration(TimeUnit::Nanosecond) => {
+                typed_min_max_batch!(
+                    $VALUES,
+                    DurationNanosecondArray,
+                    DurationNanosecond,
+                    $OP
+                )
+            }
             other => {
                 // This should have been handled before
                 return internal_err!(
@@ -1597,7 +1625,7 @@ mod tests {
         assert_eq!(
             min_res,
             ScalarValue::IntervalYearMonth(Some(IntervalYearMonthType::make_value(
-                -2, 4
+                -2, 4,
             )))
         );
 
@@ -1609,7 +1637,7 @@ mod tests {
         assert_eq!(
             max_res,
             ScalarValue::IntervalYearMonth(Some(IntervalYearMonthType::make_value(
-                5, 34
+                5, 34,
             )))
         );
 

datafusion/sqllogictest/test_files/aggregate.slt

@@ -3807,6 +3807,38 @@ SELECT MIN(value), MAX(value) FROM timestampmicrosecond
 statement ok
 DROP TABLE timestampmicrosecond;
 
+# min_duration, max_duration
+statement ok
+create table d
+as values
+  (arrow_cast(1, 'Duration(Second)'), arrow_cast(2, 'Duration(Millisecond)'),  arrow_cast(3, 'Duration(Microsecond)'),  arrow_cast(4, 'Duration(Nanosecond)'), 1),
+  (arrow_cast(11, 'Duration(Second)'),arrow_cast(22, 'Duration(Millisecond)'), arrow_cast(33, 'Duration(Microsecond)'), arrow_cast(44, 'Duration(Nanosecond)'), 1);
+
+query ????
+SELECT min(column1), min(column2), min(column3), min(column4) FROM d;
+----
+0 days 0 hours 0 mins 1 secs 0 days 0 hours 0 mins 0.002 secs 0 days 0 hours 0 mins 0.000003 secs 0 days 0 hours 0 mins 0.000000004 secs
+
+query ????
+SELECT max(column1), max(column2), max(column3), max(column4) FROM d;
+----
+0 days 0 hours 0 mins 11 secs 0 days 0 hours 0 mins 0.022 secs 0 days 0 hours 0 mins 0.000033 secs 0 days 0 hours 0 mins 0.000000044 secs
+
+# GROUP BY follows a different code path
+query ????I
+SELECT min(column1), min(column2), min(column3), min(column4), column5 FROM d GROUP BY column5;
+----
+0 days 0 hours 0 mins 1 secs 0 days 0 hours 0 mins 0.002 secs 0 days 0 hours 0 mins 0.000003 secs 0 days 0 hours 0 mins 0.000000004 secs 1
+
+query ????I
+SELECT max(column1), max(column2), max(column3), max(column4), column5 FROM d GROUP BY column5;
+----
+0 days 0 hours 0 mins 11 secs 0 days 0 hours 0 mins 0.022 secs 0 days 0 hours 0 mins 0.000033 secs 0 days 0 hours 0 mins 0.000000044 secs 1
+
+
+statement ok
+drop table d;
+
 # max_bool
 statement ok
 CREATE TABLE max_bool (value BOOLEAN);