lucasfang commented on code in PR #388:
URL: https://github.com/apache/paimon-cpp/pull/388#discussion_r4091200471
##########
src/paimon/common/data/blob_utils.cpp:
##########
@@ -108,6 +112,16 @@ bool BlobUtils::IsBlobField(const
std::shared_ptr<arrow::Field>& field) {
return IsBlobMetadata(field->metadata());
}
+bool BlobUtils::IsArrayBlobField(const std::shared_ptr<arrow::Field>& field) {
+ if (field == nullptr || field->type()->id() != arrow::Type::LIST) {
+ return false;
+ }
+ const auto& list_type = checked_cast<const
arrow::ListType&>(*field->type());
+ // Arrow's C schema importer passes MakeChildField(0) directly to
ListType, retaining the
+ // element field's metadata.
+ return IsBlobField(list_type.value_field());
Review Comment:
IsArrayBlobField / IsMapBlobField use inconsistent detection strategies —
src/paimon/common/data/blob_utils.cppIsMapBlobField deliberately avoids relying
on nested-field metadata ("Arrow's C schema bridge does not retain nested field
metadata for MapType") and instead does a structural check:
map_type.item_type()->id() == LARGE_BINARY, justified by the fact that
"Paimon's ordinary binary type is BINARY, so LARGE_BINARY uniquely identifies a
BLOB value here."The new IsArrayBlobField takes the opposite approach and
depends entirely on the element field's metadata surviving the bridge: return
IsBlobField(list_type.value_field());. So the same "container BLOB" concept is
detected structurally for MAP but via metadata for ARRAY.
--
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]
