feat: Remove use of nightly int_roundings (apache#228)

core/src/execution/datafusion/expressions/avg_decimal.rs

@@ -34,7 +34,7 @@ use arrow_data::decimal::{
     validate_decimal_precision, MAX_DECIMAL_FOR_EACH_PRECISION, MIN_DECIMAL_FOR_EACH_PRECISION,
 };
 
-use num::Integer;
+use num::{integer::div_ceil, Integer};
 use DataType::*;
 
 /// AVG aggregate expression
@@ -514,7 +514,7 @@ fn avg(sum: i128, count: i128, target_min: i128, target_max: i128, scaler: i128)
     if let Some(value) = sum.checked_mul(scaler) {
         // `sum / count` with ROUND_HALF_UP
         let (div, rem) = value.div_rem(&count);
-        let half = count.div_ceil(2);
+        let half = div_ceil(count, 2);
         let half_neg = half.neg_wrapping();
         let new_value = match value >= 0 {
             true if rem >= half => div.add_wrapping(1),

core/src/execution/datafusion/expressions/scalar_funcs.rs

@@ -37,7 +37,10 @@ use datafusion_common::{
 use datafusion_physical_expr::{
     execution_props::ExecutionProps, functions::create_physical_fun, math_expressions,
 };
-use num::{BigInt, Signed, ToPrimitive};
+use num::{
+    integer::{div_ceil, div_floor},
+    BigInt, Signed, ToPrimitive,
+};
 use unicode_segmentation::UnicodeSegmentation;
 
 /// Create a physical scalar function.
@@ -249,13 +252,13 @@ fn long_to_decimal(v: &Option<i64>, precision: u8) -> Option<i128> {
 #[inline]
 fn decimal_ceil_f(scale: &i8) -> impl Fn(i128) -> i128 {
     let div = 10_i128.pow_wrapping(*scale as u32);
-    move |x: i128| x.div_ceil(div)
+    move |x: i128| div_ceil(x, div)
 }
 
 #[inline]
 fn decimal_floor_f(scale: &i8) -> impl Fn(i128) -> i128 {
     let div = 10_i128.pow_wrapping(*scale as u32);
-    move |x: i128| x.div_floor(div)
+    move |x: i128| div_floor(x, div)
 }
 
 // Spark uses BigDecimal. See RoundBase implementation in Spark. Instead, we do the same by

core/src/execution/datafusion/expressions/utils.rs

@@ -31,6 +31,7 @@ use arrow_schema::DataType;
 use chrono::{DateTime, Offset, TimeZone};
 use datafusion_common::cast::as_generic_string_array;
 use datafusion_physical_expr::PhysicalExpr;
+use num::integer::div_floor;
 use std::{any::Any, sync::Arc};
 
 /// An utility function from DataFusion. It is not exposed by DataFusion.
@@ -198,13 +199,13 @@ pub(crate) fn spark_cast(array: ArrayRef, from_type: &DataType, to_type: &DataTy
     match (from_type, to_type) {
         (DataType::Timestamp(_, _), DataType::Int64) => {
             // See Spark's `Cast` expression
-            unary_dyn::<_, Int64Type>(&array, |v| v.div_floor(MICROS_PER_SECOND)).unwrap()
+            unary_dyn::<_, Int64Type>(&array, |v| div_floor(v, MICROS_PER_SECOND)).unwrap()
         }
         (DataType::Dictionary(_, value_type), DataType::Int64)
             if matches!(value_type.as_ref(), &DataType::Timestamp(_, _)) =>
         {
             // See Spark's `Cast` expression
-            unary_dyn::<_, Int64Type>(&array, |v| v.div_floor(MICROS_PER_SECOND)).unwrap()
+            unary_dyn::<_, Int64Type>(&array, |v| div_floor(v, MICROS_PER_SECOND)).unwrap()
         }
         (DataType::Timestamp(_, _), DataType::Utf8) => remove_trailing_zeroes(array),
         (DataType::Dictionary(_, value_type), DataType::Utf8)

core/src/lib.rs

@@ -21,7 +21,6 @@
 #![allow(clippy::upper_case_acronyms)]
 #![allow(clippy::derive_partial_eq_without_eq)] // For prost generated struct
 #![cfg_attr(feature = "nightly", feature(core_intrinsics))]
-#![feature(int_roundings)]
 #![feature(specialization)]
 
 // Branch prediction hint. This is currently only available on nightly.