ARROW-11349: [Rust] Add from_iter_values to create arrays from (non n…

…ull) values

The idea of this PR is to have a function `from_iter_values` that (just like `from_iter`) creates an array based on an iterator, but from `T` instead of `Option<T>`.

I have seen some places in DataFusion (especially `to_array_of_size`) where an `Array` is generated from a `Vec` of items, which could be replaced by this.
The other iterators have some memory / time overhead in both creating and manipulating the null buffer (and in the case of `Vec` for allocating / dropping the Vec)

Closes apache#9293 from Dandandan/array_iter_non_null

Authored-by: Heres, Daniel <danielheres@gmail.com>
Signed-off-by: Jorge C. Leitao <jorgecarleitao@gmail.com>

rust/arrow/src/array/array_primitive.rs

@@ -94,6 +94,21 @@ impl<T: ArrowPrimitiveType> PrimitiveArray<T> {
         let offset = i + self.offset();
         unsafe { *self.raw_values.as_ptr().add(offset) }
     }
+
+    /// Creates a PrimitiveArray based on an iterator of values without nulls
+    pub fn from_iter_values<I: IntoIterator<Item = T::Native>>(iter: I) -> Self {
+        let val_buf: Buffer = iter.into_iter().collect();
+        let data = ArrayData::new(
+            T::DATA_TYPE,
+            val_buf.len() / mem::size_of::<<T as ArrowPrimitiveType>::Native>(),
+            None,
+            None,
+            0,
+            vec![val_buf],
+            vec![],
+        );
+        PrimitiveArray::from(Arc::new(data))
+    }
 }
 
 impl<T: ArrowPrimitiveType> Array for PrimitiveArray<T> {
@@ -820,6 +835,18 @@ mod tests {
         }
     }
 
+    #[test]
+    fn test_primitive_from_iter_values() {
+        // Test building a primitive array with from_iter_values
+
+        let arr: PrimitiveArray<Int32Type> = PrimitiveArray::from_iter_values(0..10);
+        assert_eq!(10, arr.len());
+        assert_eq!(0, arr.null_count());
+        for i in 0..10i32 {
+            assert_eq!(i, arr.value(i as usize));
+        }
+    }
+
     #[test]
     #[should_panic(expected = "PrimitiveArray data should contain a single buffer only \
                                (values buffer)")]

rust/arrow/src/array/array_string.rs

@@ -146,6 +146,36 @@ impl<OffsetSize: StringOffsetSizeTrait> GenericStringArray<OffsetSize> {
     pub(crate) fn from_opt_vec(v: Vec<Option<&str>>) -> Self {
         v.into_iter().collect()
     }
+
+    /// Creates a `GenericStringArray` based on an iterator of values without nulls
+    pub fn from_iter_values<Ptr, I: IntoIterator<Item = Ptr>>(iter: I) -> Self
+    where
+        Ptr: AsRef<str>,
+    {
+        let iter = iter.into_iter();
+        let (_, data_len) = iter.size_hint();
+        let data_len = data_len.expect("Iterator must be sized"); // panic if no upper bound.
+
+        let mut offsets =
+            MutableBuffer::new((data_len + 1) * std::mem::size_of::<OffsetSize>());
+        let mut values = MutableBuffer::new(0);
+
+        let mut length_so_far = OffsetSize::zero();
+        offsets.push(length_so_far);
+
+        for i in iter {
+            let s = i.as_ref();
+            length_so_far += OffsetSize::from_usize(s.len()).unwrap();
+            offsets.push(length_so_far);
+            values.extend_from_slice(s.as_bytes());
+        }
+        let array_data = ArrayData::builder(OffsetSize::DATA_TYPE)
+            .len(data_len)
+            .add_buffer(offsets.into())
+            .add_buffer(values.into())
+            .build();
+        Self::from(array_data)
+    }
 }
 
 impl<'a, Ptr, OffsetSize: StringOffsetSizeTrait> FromIterator<Option<Ptr>>
@@ -411,6 +441,7 @@ mod tests {
         );
     }
 
+    #[test]
     fn test_string_array_from_iter() {
         let data = vec![Some("hello"), None, Some("arrow")];
         // from Vec<Option<&str>>
@@ -424,4 +455,13 @@ mod tests {
         assert_eq!(array1, array2);
         assert_eq!(array2, array3);
     }
+
+    #[test]
+    fn test_string_array_from_iter_values() {
+        let data = vec!["hello", "hello2"];
+        let array1 = StringArray::from_iter_values(data.iter());
+
+        assert_eq!(array1.value(0), "hello");
+        assert_eq!(array1.value(1), "hello2");
+    }
 }