lidavidm commented on a change in pull request #10397:
URL: https://github.com/apache/arrow/pull/10397#discussion_r640905731
##########
File path: cpp/src/arrow/compute/exec/expression.cc
##########
@@ -492,9 +448,18 @@ Result<Expression> Expression::Bind(ValueDescr in,
if (literal()) return *this;
if (auto ref = field_ref()) {
- ARROW_ASSIGN_OR_RAISE(auto field, ref->GetOneOrNone(*in.type));
- auto descr = field ? ValueDescr{field->type(), in.shape} :
ValueDescr::Scalar(null());
- return Expression{Parameter{*ref, std::move(descr)}};
+ if (ref->IsNested()) {
+ return Status::NotImplemented("nested field references");
+ }
+
+ ARROW_ASSIGN_OR_RAISE(auto path, ref->FindOne(*in.type));
Review comment:
Yeah, when I was digging into the wide-dataframe performance regression,
it boiled almost entirely down to us doing a linear field lookup for every
field, for every batch (ARROW-11469)
--
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.
For queries about this service, please contact Infrastructure at:
[email protected]
