Rich-T-kid commented on code in PR #10812:
URL: https://github.com/apache/arrow-rs/pull/10812#discussion_r3873551268
##########
arrow-select/src/take.rs:
##########
@@ -652,76 +651,142 @@ where
OffsetType::Native: OffsetSizeTrait,
PrimitiveArray<OffsetType>: From<Vec<OffsetType::Native>>,
{
- let list_offsets = values.value_offsets();
+ let src_offsets = values.value_offsets();
let child_data = values.values().to_data();
let nulls = take_nulls(values.nulls(), indices);
- let mut new_offsets = Vec::with_capacity(indices.len() + 1);
- new_offsets.push(OffsetType::Native::zero());
+ let mut dst_offsets = Vec::with_capacity(indices.len() + 1);
+ dst_offsets.push(OffsetType::Native::zero());
- let use_nulls = child_data.null_count() > 0;
+ let field = values.value_field().clone();
+
+ let is_primitive_child = child_data.null_count() == 0 &&
child_data.data_type().is_primitive();
+
+ if is_primitive_child {
+ let values_buf = &child_data.buffers()[0];
+ let bytes_per_value = if !child_data.is_empty() {
+ values_buf.len() / child_data.len()
+ } else {
+ 0
+ };
+ let child_buf_offset = child_data.offset() * bytes_per_value;
+
+ let avg_row_len = child_data
+ .len()
+ .checked_div(values.len().max(1))
+ .unwrap_or(0);
+ let mut dst_buf = MutableBuffer::new(
+ avg_row_len
+ .saturating_mul(indices.len())
+ .saturating_mul(bytes_per_value),
+ );
+
+ let mut child_len = OffsetType::Native::zero();
+
+ match nulls.as_ref().filter(|n| n.null_count() > 0) {
+ None => {
+ for &idx in indices.values() {
+ let row = idx.as_usize();
+ let start = child_buf_offset + src_offsets[row].as_usize()
* bytes_per_value;
+ let end = child_buf_offset + src_offsets[row +
1].as_usize() * bytes_per_value;
+ dst_buf.extend_from_slice(&values_buf[start..end]);
+ child_len += src_offsets[row + 1] - src_offsets[row];
Review Comment:
added a `checked_add()` here with some error handling
--
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]
