fix: handle dictionary-encoded timestamps and add TimestampNTZ regression test

- Add support for dictionary-encoded timestamps in extract_date_part
- Add comprehensive test for hour/minute/second with TimestampNTZ in non-UTC timezones
- Addresses reviewer feedback on PR #3265 for issue #3180

Author: vigneshsiva11

native/spark-expr/src/datetime_funcs/extract_date_part.rs

@@ -76,9 +76,20 @@ macro_rules! extract_date_part {
 
                 match args {
                     [ColumnarValue::Array(array)] => {
+                        // First, normalize dictionary-encoded arrays (common in Parquet/Iceberg)
+                        let array = match array.data_type() {
+                            DataType::Dictionary(_, value_type) => {
+                                // Cast dictionary to the underlying timestamp type
+                                arrow::compute::cast(&array, value_type.as_ref())
+                                    .map_err(|e| DataFusionError::Execution(e.to_string()))?
+                            }
+                            _ => array.clone(),
+                        };
+
+                        // Then handle timezone conversion based on timestamp type
                         let array = match array.data_type() {
                             // TimestampNTZ → DO NOT apply timezone conversion
-                            DataType::Timestamp(_, None) => array.clone(),
+                            DataType::Timestamp(_, None) => array,
 
                             // Timestamp with timezone → convert from UTC to session timezone
                             DataType::Timestamp(_, Some(_)) => array_with_timezone(

spark/src/test/scala/org/apache/comet/CometTemporalExpressionSuite.scala

@@ -395,4 +395,30 @@ class CometTemporalExpressionSuite extends CometTestBase with AdaptiveSparkPlanH
     // Test null handling
     checkSparkAnswerAndOperator("SELECT unix_date(NULL)")
   }
+
+  test("hour/minute/second with TimestampNTZ in non-UTC timezone") {
+    // Regression test for issue #3180
+    // TimestampNTZ stores local time without timezone information
+    // hour/minute/second should extract directly from local time without timezone conversion
+    val schema = StructType(Seq(StructField("ts_ntz", DataTypes.TimestampNTZType, true)))
+    
+    // Create test data with known TimestampNTZ values
+    val data = Seq(
+      Row(java.time.LocalDateTime.of(2024, 1, 15, 10, 30, 45)),  // 10:30:45
+      Row(java.time.LocalDateTime.of(2024, 6, 20, 14, 15, 20)),  // 14:15:20
+      Row(java.time.LocalDateTime.of(2024, 12, 31, 23, 59, 59)), // 23:59:59
+      Row(null))
+    
+    val df = spark.createDataFrame(spark.sparkContext.parallelize(data), schema)
+    df.createOrReplaceTempView("timestamp_ntz_tbl")
+
+    // Test in multiple timezones - results should be the same since TimestampNTZ has no timezone
+    for (timezone <- Seq("UTC", "America/Los_Angeles", "Asia/Tokyo")) {
+      withSQLConf(SQLConf.SESSION_LOCAL_TIMEZONE.key -> timezone) {
+        // hour() should return the hour from local time directly
+        checkSparkAnswerAndOperator(
+          "SELECT ts_ntz, hour(ts_ntz), minute(ts_ntz), second(ts_ntz) FROM timestamp_ntz_tbl ORDER BY ts_ntz")
+      }
+    }
+  }
 }