pitrou commented on code in PR #50976:
URL: https://github.com/apache/arrow/pull/50976#discussion_r4008366588
##########
cpp/src/arrow/compute/kernels/scalar_cast_nested.cc:
##########
@@ -141,6 +142,138 @@ void AddListCast(CastFunction* func) {
DCHECK_OK(func->AddKernel(SrcType::type_id, std::move(kernel)));
}
+template <typename SrcType, typename DestType>
+struct CastListView {
+ using src_offset_type = typename SrcType::offset_type;
+ using dest_offset_type = typename DestType::offset_type;
+
+ static constexpr bool is_downcast = sizeof(src_offset_type) >
sizeof(dest_offset_type);
+
+ static bool IsContiguous(const ArraySpan& in_array) {
+ const auto* offsets = in_array.GetValues<src_offset_type>(1);
+ const auto* sizes = in_array.GetValues<src_offset_type>(2);
+ for (int64_t i = 0; i < in_array.length - 1; ++i) {
+ if (in_array.IsNull(i) && sizes[i] != 0) {
+ return false;
+ }
+ if (offsets[i] + sizes[i] != offsets[i + 1]) {
+ return false;
+ }
+ }
+ if (in_array.length > 0 && in_array.IsNull(in_array.length - 1) &&
+ sizes[in_array.length - 1] != 0) {
+ return false;
+ }
+ return true;
+ }
+
+ static Status Exec(KernelContext* ctx, const ExecSpan& batch, ExecResult*
out) {
+ const CastOptions& options = CastState::Get(ctx);
+ auto child_type = checked_cast<const DestType&>(*out->type()).value_type();
+ const ArraySpan& in_array = batch[0].array;
+ ArrayData* out_array = out->array_data().get();
+
+ if (in_array.length == 0) {
+ out_array->buffers[0] = nullptr;
+ ARROW_ASSIGN_OR_RAISE(out_array->buffers[1],
+ ctx->Allocate(sizeof(dest_offset_type)));
+ auto* dest_offsets = out_array->GetMutableValues<dest_offset_type>(1);
+ dest_offsets[0] = 0;
+ std::shared_ptr<ArrayData> values = in_array.child_data[0].ToArrayData();
+ ARROW_ASSIGN_OR_RAISE(Datum cast_values, Cast(values->Slice(0, 0),
child_type,
+ options,
ctx->exec_context()));
+ DCHECK(cast_values.is_array());
+ out_array->child_data.push_back(cast_values.array());
+ return Status::OK();
+ }
+
+ ARROW_ASSIGN_OR_RAISE(out_array->buffers[0],
+ GetOrCopyNullBitmapBuffer(in_array,
ctx->memory_pool()));
+
+ std::shared_ptr<ArrayData> values = in_array.child_data[0].ToArrayData();
+
+ const auto* offsets = in_array.GetValues<src_offset_type>(1);
+ const auto* sizes = in_array.GetValues<src_offset_type>(2);
+
+ if (IsContiguous(in_array)) {
+ // Zero-copy fast-path: shift offsets and slice child values
+ ARROW_ASSIGN_OR_RAISE(
+ out_array->buffers[1],
+ ctx->Allocate(sizeof(dest_offset_type) * (in_array.length + 1)));
+ auto* dest_offsets = out_array->GetMutableValues<dest_offset_type>(1);
+
+ src_offset_type start_offset = offsets[0];
+ src_offset_type end_offset =
+ offsets[in_array.length - 1] + sizes[in_array.length - 1] -
start_offset;
+
+ if constexpr (is_downcast) {
+ if (end_offset > std::numeric_limits<dest_offset_type>::max()) {
+ return Status::Invalid("Array of type ", in_array.type->ToString(),
+ " too large to convert to ",
+ out_array->type->ToString());
+ }
+ }
+
+ for (int64_t i = 0; i < in_array.length; ++i) {
+ dest_offsets[i] = static_cast<dest_offset_type>(offsets[i] -
start_offset);
+ }
+ dest_offsets[in_array.length] =
static_cast<dest_offset_type>(end_offset);
+
+ values = values->Slice(start_offset, dest_offsets[in_array.length]);
+ } else {
+ // Non-contiguous path: compute new offsets, flatten/concatenate values
+ ARROW_ASSIGN_OR_RAISE(
+ out_array->buffers[1],
+ ctx->Allocate(sizeof(dest_offset_type) * (in_array.length + 1)));
+ auto* dest_offsets = out_array->GetMutableValues<dest_offset_type>(1);
+
+ src_offset_type current_offset = 0;
+ dest_offsets[0] = 0;
+ for (int64_t i = 0; i < in_array.length; ++i) {
+ if (in_array.IsNull(i)) {
+ dest_offsets[i + 1] = static_cast<dest_offset_type>(current_offset);
+ } else {
+ current_offset += sizes[i];
+ dest_offsets[i + 1] = static_cast<dest_offset_type>(current_offset);
+ }
+ }
Review Comment:
This part wasn't addressed:
> That loop would compute all destination offsets and compute whether the
source entries are contiguous, all in one go.
i.e. try to fold the contiguity check (`IsContiguous`) into the existing
for-loop that also computes destination offsets.
--
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]
