jorisvandenbossche commented on code in PR #40803:
URL: https://github.com/apache/arrow/pull/40803#discussion_r1540665821
##########
cpp/src/arrow/record_batch.cc:
##########
@@ -334,42 +347,51 @@ Result<std::shared_ptr<Tensor>>
RecordBatch::ToTensor(MemoryPool* pool) const {
result_type = result_field->type();
}
+ // Check if result_type is signed or unsigned integer and null_to_nan is set
to true
+ // Then all columns should be promoted to float type
+ if (is_integer(result_type->id()) && null_to_nan) {
+ ARROW_ASSIGN_OR_RAISE(
+ result_field,
+ result_field->MergeWith(field(result_field->name(), float16()),
options));
Review Comment:
Ah, but of course you can also write the test without float16 in the input,
and just have int8 that need to be upcasted to float32 and not float16. So on
your
> Should I also add int8 with nulls?
Yes, let's do that. As that is the case that would have been silently giving
garbage results when using float16 instead of float32
--
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]
