lidavidm commented on a change in pull request #10530:
URL: https://github.com/apache/arrow/pull/10530#discussion_r657032193
##########
File path: cpp/src/arrow/compute/kernels/scalar_arithmetic.cc
##########
@@ -454,6 +454,114 @@ struct PowerChecked {
}
};
+// Bitwise operations
+
+struct BitWiseNot {
+ template <typename T, typename Arg>
+ static T Call(KernelContext*, Arg arg, Status*) {
+ return ~arg;
+ }
+};
+
+struct BitWiseAnd {
+ template <typename T, typename Arg0, typename Arg1>
+ static T Call(KernelContext*, Arg0 lhs, Arg1 rhs, Status*) {
+ return lhs & rhs;
+ }
+};
+
+struct BitWiseOr {
+ template <typename T, typename Arg0, typename Arg1>
+ static T Call(KernelContext*, Arg0 lhs, Arg1 rhs, Status*) {
+ return lhs | rhs;
+ }
+};
+
+struct BitWiseXor {
+ template <typename T, typename Arg0, typename Arg1>
+ static T Call(KernelContext*, Arg0 lhs, Arg1 rhs, Status*) {
+ return lhs ^ rhs;
+ }
+};
+
+struct ShiftLeft {
+ template <typename T, typename Arg0, typename Arg1>
+ static T Call(KernelContext*, Arg0 lhs, Arg1 rhs, Status*) {
+ using Unsigned = typename std::make_unsigned<Arg0>::type;
+ static_assert(std::is_same<T, Arg0>::value, "");
+ if (ARROW_PREDICT_FALSE(rhs < 0 || rhs >=
std::numeric_limits<Arg0>::digits)) {
+ return lhs;
+ }
+ return static_cast<T>(static_cast<Unsigned>(lhs) <<
static_cast<Unsigned>(rhs));
+ }
+};
+
+// See SEI CERT C Coding Standard rule INT34-C
+struct ShiftLeftChecked {
+ template <typename T, typename Arg0, typename Arg1>
+ static enable_if_unsigned_integer<T> Call(KernelContext*, Arg0 lhs, Arg1 rhs,
+ Status* st) {
+ static_assert(std::is_same<T, Arg0>::value, "");
+ if (ARROW_PREDICT_FALSE(rhs < 0)) {
+ *st = Status::Invalid("shift must be non-negative");
+ return lhs;
+ }
+ if (ARROW_PREDICT_FALSE(rhs >= std::numeric_limits<Arg0>::digits)) {
+ *st = Status::Invalid("overflow");
+ return lhs;
+ }
+ return lhs << rhs;
Review comment:
I thought for shifts, usually it's desired to take the remainder if
there would be overflow? Though [Wikipedia actually differentiates between
arithmetic and logical left
shift](https://en.wikipedia.org/wiki/Arithmetic_shift#Equivalence_of_arithmetic_and_logical_left_shifts_and_multiplication)
here: the former raises overflow errors, the latter returns the modulo.
(Also, I think it's just sufficient to check that (lhs << rhs) < lhs.)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
