etseidl commented on code in PR #37940:
URL: https://github.com/apache/arrow/pull/37940#discussion_r1342890135
##########
cpp/src/parquet/encoding.cc:
##########
@@ -2250,17 +2253,17 @@ void DeltaBitPackEncoder<DType>::FlushBlock() {
std::min(values_per_mini_block_, values_current_block_);
const uint32_t start = i * values_per_mini_block_;
- const UT max_delta = *std::max_element(
+ const T max_delta = *std::max_element(
deltas_.begin() + start, deltas_.begin() + start +
values_current_mini_block);
// The minimum number of bits required to write any of values in deltas_
vector.
// See overflow comment above.
- const auto bit_width = bit_width_data[i] =
- bit_util::NumRequiredBits(max_delta - min_delta);
+ const auto bit_width = bit_width_data[i] = bit_util::NumRequiredBits(
+ static_cast<UT>(SafeSignedSubtract(max_delta, min_delta)));
Review Comment:
I don't think we would want that. It's ok for the subtraction to wrap around
and become negative, we just cast it back to unsigned. On the decoding side it
will wrap around again and yield the correct result. The only reason to use
`SafeSignedSubtract` is to get well defined wrapping behavior, which signed
subtraction does not give us.
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
