Support dictionary data type in array_to_string (apache#10908)

* Support dictionary data type in array_to_string

* Fix import

* Some tests

* Update datafusion/functions-array/src/string.rs

Co-authored-by: Alex Huang <huangweijun1001@gmail.com>

* Add some tests showing incorrect results

* Get logical array

* apply rust fmt

* Simplify implementation, avoid panics

---------

Co-authored-by: Alex Huang <huangweijun1001@gmail.com>
Co-authored-by: Andrew Lamb <andrew@nerdnetworks.org>

Author: 3 people

datafusion/functions-array/src/string.rs

@@ -26,12 +26,15 @@ use arrow::array::{
 use arrow::datatypes::{DataType, Field};
 use datafusion_expr::TypeSignature;
 
-use datafusion_common::{plan_err, DataFusionError, Result};
+use datafusion_common::{not_impl_err, plan_err, DataFusionError, Result};
 
 use std::any::{type_name, Any};
 
 use crate::utils::{downcast_arg, make_scalar_function};
-use arrow_schema::DataType::{FixedSizeList, LargeList, LargeUtf8, List, Null, Utf8};
+use arrow::compute::cast;
+use arrow_schema::DataType::{
+    Dictionary, FixedSizeList, LargeList, LargeUtf8, List, Null, Utf8,
+};
 use datafusion_common::cast::{
     as_generic_string_array, as_large_list_array, as_list_array, as_string_array,
 };
@@ -76,7 +79,7 @@ macro_rules! call_array_function {
             DataType::UInt16 => array_function!(UInt16Array),
             DataType::UInt32 => array_function!(UInt32Array),
             DataType::UInt64 => array_function!(UInt64Array),
-            _ => unreachable!(),
+            dt => not_impl_err!("Unsupported data type in array_to_string: {dt}"),
         }
     };
     ($DATATYPE:expr, $INCLUDE_LIST:expr) => {{
@@ -95,7 +98,7 @@ macro_rules! call_array_function {
             DataType::UInt16 => array_function!(UInt16Array),
             DataType::UInt32 => array_function!(UInt32Array),
             DataType::UInt64 => array_function!(UInt64Array),
-            _ => unreachable!(),
+            dt => not_impl_err!("Unsupported data type in array_to_string: {dt}"),
         }
     }};
 }
@@ -245,6 +248,8 @@ pub(super) fn array_to_string_inner(args: &[ArrayRef]) -> Result<ArrayRef> {
         with_null_string = true;
     }
 
+    /// Creates a single string from single element of a ListArray (which is
+    /// itself another Array)
     fn compute_array_to_string(
         arg: &mut String,
         arr: ArrayRef,
@@ -281,6 +286,22 @@ pub(super) fn array_to_string_inner(args: &[ArrayRef]) -> Result<ArrayRef> {
 
                 Ok(arg)
             }
+            Dictionary(_key_type, value_type) => {
+                // Call cast to unwrap the dictionary. This could be optimized if we wanted
+                // to accept the overhead of extra code
+                let values = cast(&arr, value_type.as_ref()).map_err(|e| {
+                    DataFusionError::from(e).context(
+                        "Casting dictionary to values in compute_array_to_string",
+                    )
+                })?;
+                compute_array_to_string(
+                    arg,
+                    values,
+                    delimiter,
+                    null_string,
+                    with_null_string,
+                )
+            }
             Null => Ok(arg),
             data_type => {
                 macro_rules! array_function {

datafusion/sqllogictest/test_files/array.slt

@@ -3769,6 +3769,54 @@ select array_to_string(make_array(), ',')
 ----
 (empty)
 
+# array to string dictionary
+statement ok
+CREATE TABLE table1 AS VALUES
+  (1, 'foo'),
+  (3, 'bar'),
+  (1, 'foo'),
+  (2, NULL),
+  (NULL, 'baz')
+  ;
+
+# expect 1-3-1-2 (dictionary values should be repeated)
+query T
+SELECT array_to_string(array_agg(column1),'-')
+FROM (
+  SELECT arrow_cast(column1, 'Dictionary(Int32, Int32)') as column1
+  FROM table1
+);
+----
+1-3-1-2
+
+# expect foo,bar,foo,baz (dictionary values should be repeated)
+query T
+SELECT array_to_string(array_agg(column2),',')
+FROM (
+  SELECT arrow_cast(column2, 'Dictionary(Int64, Utf8)') as column2
+  FROM table1
+);
+----
+foo,bar,foo,baz
+
+# Expect only values that are in the group
+query I?T
+SELECT column1, array_agg(column2), array_to_string(array_agg(column2),',')
+FROM (
+  SELECT column1, arrow_cast(column2, 'Dictionary(Int32, Utf8)') as column2
+  FROM table1
+)
+GROUP BY column1
+ORDER BY column1;
+----
+1 [foo, foo] foo,foo
+2 [] (empty)
+3 [bar] bar
+NULL [baz] baz
+
+statement ok
+drop table table1;
+
 
 ## array_union (aliases: `list_union`)