sdf-jkl commented on code in PR #10313:
URL: https://github.com/apache/arrow-rs/pull/10313#discussion_r3731631478
##########
parquet-variant-compute/src/variant_to_arrow.rs:
##########
@@ -661,6 +674,216 @@ impl<'a> StructVariantToArrowRowBuilder<'a> {
}
}
+/// Builder for converting variant values into a [`UnionArray`].
+///
+/// Each value is dispatched to the union field that most exactly represents
its runtime type
+/// (see [`union_child_rank`]), with ties broken by declaration order. Unions
have no top-level
+/// null buffer, so null rows -- and, in safe mode, values no field can
represent -- become a
+/// null in the [`DataType::Null`] child if the union declares one, otherwise
in the first child.
+pub(crate) struct UnionVariantToArrowRowBuilder<'a> {
+ fields: &'a UnionFields,
+ mode: UnionMode,
+ children: Vec<UnionChildBuilder<'a>>,
+ type_ids: Vec<i8>,
+ /// Dense mode only
+ offsets: Vec<i32>,
+ null_child: usize,
+ cast_options: &'a CastOptions<'a>,
+}
+
+struct UnionChildBuilder<'a> {
+ type_id: i8,
+ builder: VariantToArrowRowBuilder<'a>,
+ len: i32,
+}
+
+impl<'a> UnionVariantToArrowRowBuilder<'a> {
+ fn try_new(
+ fields: &'a UnionFields,
+ mode: UnionMode,
+ cast_options: &'a CastOptions<'a>,
+ capacity: usize,
+ ) -> Result<Self> {
+ // null rows need a child to land in
+ if fields.is_empty() {
+ return Err(ArrowError::InvalidArgumentError(
+ "Casting Variant to a union requires at least one union
field".to_string(),
+ ));
+ }
+ let mut children = Vec::with_capacity(fields.len());
+ for (type_id, field) in fields.iter() {
+ children.push(UnionChildBuilder {
+ type_id,
+ builder: make_typed_variant_to_arrow_row_builder(
+ field.data_type(),
+ cast_options,
+ capacity,
+ )?,
+ len: 0,
+ });
+ }
+ let null_child = fields
+ .iter()
+ .position(|(_, field)| field.data_type() == &DataType::Null)
+ .unwrap_or(0);
Review Comment:
`UnionArray`s are confusing :weary: They don't seem to enforce nullability
of the child arrays, so it's fine to append Null to any array if a Null array
is not present
--
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]
