fix(dataframe): handle FixedSizeBinary in describe

Skip min/max describe summaries for unsupported binary-like types so describe falls back to nulls instead of attempting an invalid Utf8 cast. Add a regression test for FixedSizeBinary and rerun the dataframe describe integration tests reported in #20273.

Author: officialasishkumar

datafusion/core/src/dataframe/mod.rs

@@ -1015,6 +1015,16 @@ impl DataFrame {
         //the functions now supported
         let supported_describe_functions =
             vec!["count", "null_count", "mean", "std", "min", "max", "median"];
+        let supports_describe_min_max = |data_type: &DataType| {
+            !matches!(
+                data_type,
+                DataType::Boolean
+                    | DataType::Binary
+                    | DataType::LargeBinary
+                    | DataType::BinaryView
+                    | DataType::FixedSizeBinary(_)
+            )
+        };
 
         let original_schema_fields = self.schema().fields().iter();
 
@@ -1075,9 +1085,7 @@ impl DataFrame {
                 vec![],
                 original_schema_fields
                     .clone()
-                    .filter(|f| {
-                        !matches!(f.data_type(), DataType::Binary | DataType::Boolean)
-                    })
+                    .filter(|f| supports_describe_min_max(f.data_type()))
                     .map(|f| min(ident(f.name())).alias(f.name()))
                     .collect::<Vec<_>>(),
             ),
@@ -1086,9 +1094,7 @@ impl DataFrame {
                 vec![],
                 original_schema_fields
                     .clone()
-                    .filter(|f| {
-                        !matches!(f.data_type(), DataType::Binary | DataType::Boolean)
-                    })
+                    .filter(|f| supports_describe_min_max(f.data_type()))
                     .map(|f| max(ident(f.name())).alias(f.name()))
                     .collect::<Vec<_>>(),
             ),

datafusion/core/tests/dataframe/describe.rs

@@ -19,6 +19,10 @@ use datafusion::prelude::{ParquetReadOptions, SessionContext};
 use datafusion_common::test_util::batches_to_string;
 use datafusion_common::{Result, test_util::parquet_test_data};
 use insta::assert_snapshot;
+use std::sync::Arc;
+
+use arrow::array::{FixedSizeBinaryArray, RecordBatch};
+use arrow::datatypes::{DataType, Field, Schema};
 
 #[tokio::test]
 async fn describe() -> Result<()> {
@@ -81,6 +85,44 @@ async fn describe_boolean_binary() -> Result<()> {
     Ok(())
 }
 
+#[tokio::test]
+async fn describe_fixed_size_binary() -> Result<()> {
+    let ctx = SessionContext::new();
+    let batch = RecordBatch::try_new(
+        Arc::new(Schema::new(vec![Field::new(
+            "fsb",
+            DataType::FixedSizeBinary(3),
+            true,
+        )])),
+        vec![Arc::new(FixedSizeBinaryArray::from(vec![
+            Some(&[1_u8, 2, 3][..]),
+            None,
+            Some(&[4_u8, 5, 6][..]),
+        ]))],
+    )?;
+    ctx.register_batch("test", batch)?;
+
+    let result = ctx.table("test").await?.describe().await?.collect().await?;
+
+    assert_snapshot!(
+        batches_to_string(&result),
+        @r"
+    +------------+------+
+    | describe   | fsb  |
+    +------------+------+
+    | count      | 2    |
+    | null_count | 1    |
+    | mean       | null |
+    | std        | null |
+    | min        | null |
+    | max        | null |
+    | median     | null |
+    +------------+------+
+    "
+    );
+    Ok(())
+}
+
 #[tokio::test]
 async fn describe_null() -> Result<()> {
     let ctx = parquet_context().await;