pitrou commented on code in PR #37100:
URL: https://github.com/apache/arrow/pull/37100#discussion_r1365035456
##########
cpp/src/arrow/compute/kernels/aggregate_basic_internal.h:
##########
@@ -912,6 +912,122 @@ struct NullMinMaxImpl : public ScalarAggregator {
}
};
+template <SimdLevel::type SimdLevel>
+struct DictionaryMinMaxImpl : public ScalarAggregator {
+ using ThisType = DictionaryMinMaxImpl<SimdLevel>;
+
+ DictionaryMinMaxImpl(std::shared_ptr<DataType> out_type,
ScalarAggregateOptions options)
+ : options(std::move(options)),
+ out_type(std::move(out_type)),
+ has_nulls(false),
+ count(0),
+ min(nullptr),
+ max(nullptr) {
+ this->options.min_count = std::max<uint32_t>(1, this->options.min_count);
+ }
+
+ Status Consume(KernelContext* ctx, const ExecSpan& batch) override {
+ if (batch[0].is_scalar()) {
+ return Status::NotImplemented("No min/max implemented for
DictionaryScalar");
+ }
+
+ DictionaryArray dict_arr(batch[0].array.ToArrayData());
+ ARROW_ASSIGN_OR_RAISE(auto compacted_arr,
dict_arr.Compact(ctx->memory_pool()));
+ const DictionaryArray& compacted_dict_arr =
+ checked_cast<const DictionaryArray&>(*compacted_arr);
+ if (compacted_dict_arr.length() - compacted_dict_arr.null_count() == 0) {
+ return Status::OK();
+ }
+ this->has_nulls |= compacted_dict_arr.null_count() > 0;
+ this->count += compacted_dict_arr.length() -
compacted_dict_arr.null_count();
Review Comment:
> So it's a feature that DictionaryArray::null_count() only counts the
number of nulls among the dict indices?
Yes.
> Do we want null_count() return the same number as the dictionary array is
decoded.
That would be disruptive as a lot of code probably relies on the current
behavior.
--
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]
