alamb commented on code in PR #8006:
URL: https://github.com/apache/arrow-datafusion/pull/8006#discussion_r1378897737
##########
datafusion/core/src/physical_optimizer/sort_pushdown.rs:
##########
@@ -298,34 +302,34 @@ fn pushdown_requirement_to_children(
/// If the the parent requirements are more specific, push down the parent
requirements
/// If they are not compatible, need to add Sort.
fn determine_children_requirement(
- parent_required: Option<&[PhysicalSortRequirement]>,
- request_child: Option<&[PhysicalSortRequirement]>,
+ parent_required: LexRequirementRef,
Review Comment:
I almost wonder if `LexRequirement` / `LexRequirementRef` can eventually get
the same treatment (encapsulate in a struct) -- definitely not for this PR,
though ;)
##########
datafusion/physical-expr/src/physical_expr.rs:
##########
@@ -226,17 +228,78 @@ pub fn physical_exprs_contains(
.any(|physical_expr| physical_expr.eq(expr))
}
+/// Checks whether the given slices have any common entries.
+pub fn have_common_entries(
+ lhs: &[Arc<dyn PhysicalExpr>],
+ rhs: &[Arc<dyn PhysicalExpr>],
+) -> bool {
+ lhs.iter().any(|expr| physical_exprs_contains(rhs, expr))
+}
+
+/// Checks whether the given physical expression slices are equal.
+pub fn physical_exprs_equal(
+ lhs: &[Arc<dyn PhysicalExpr>],
+ rhs: &[Arc<dyn PhysicalExpr>],
+) -> bool {
+ lhs.len() == rhs.len() && izip!(lhs, rhs).all(|(lhs, rhs)| lhs.eq(rhs))
+}
+
+/// Checks whether the given physical expression slices are equal in the sense
+/// of bags (multi-sets), disregarding their orderings.
+pub fn physical_exprs_bag_equal(
+ lhs: &[Arc<dyn PhysicalExpr>],
+ rhs: &[Arc<dyn PhysicalExpr>],
+) -> bool {
+ // TODO: Once we can use `HashMap`s with `Arc<dyn PhysicalExpr>`, this
+ // function should use a `HashMap` to reduce computational
complexity.
+ if lhs.len() == rhs.len() {
+ let mut rhs_vec = rhs.to_vec();
+ for expr in lhs {
+ if let Some(idx) = rhs_vec.iter().position(|e| expr.eq(e)) {
+ rhs_vec.swap_remove(idx);
+ } else {
+ return false;
+ }
+ }
+ true
+ } else {
+ false
+ }
+}
+
+/// This utility function removes duplicates from the given `exprs` vector.
+/// Note that this function does not necessarily preserve its input ordering.
+pub fn deduplicate_physical_exprs(exprs: &mut Vec<Arc<dyn PhysicalExpr>>) {
+ // TODO: Once we can use `HashSet`s with `Arc<dyn PhysicalExpr>`, this
Review Comment:
Maybe it is worth filing a ticket about this and adding a link to that
ticket in the comments
##########
datafusion/core/src/physical_optimizer/sort_pushdown.rs:
##########
@@ -127,29 +124,27 @@ pub(crate) fn pushdown_sorts(
requirements: SortPushDown,
) -> Result<Transformed<SortPushDown>> {
let plan = &requirements.plan;
- let parent_required = requirements.required_ordering.as_deref();
- const ERR_MSG: &str = "Expects parent requirement to contain something";
- let err = || plan_datafusion_err!("{}", ERR_MSG);
+ let parent_required =
requirements.required_ordering.as_deref().unwrap_or(&[]);
if let Some(sort_exec) = plan.as_any().downcast_ref::<SortExec>() {
- let mut new_plan = plan.clone();
- if !ordering_satisfy_requirement(
- plan.output_ordering(),
- parent_required,
- || plan.equivalence_properties(),
- || plan.ordering_equivalence_properties(),
- ) {
+ let new_plan = if !plan
Review Comment:
This is a really good example of how much nicer the new API is. 👨🍳 👌 --
really nice
##########
datafusion/core/src/physical_optimizer/sort_pushdown.rs:
##########
@@ -127,29 +124,27 @@ pub(crate) fn pushdown_sorts(
requirements: SortPushDown,
) -> Result<Transformed<SortPushDown>> {
let plan = &requirements.plan;
- let parent_required = requirements.required_ordering.as_deref();
- const ERR_MSG: &str = "Expects parent requirement to contain something";
- let err = || plan_datafusion_err!("{}", ERR_MSG);
+ let parent_required =
requirements.required_ordering.as_deref().unwrap_or(&[]);
if let Some(sort_exec) = plan.as_any().downcast_ref::<SortExec>() {
- let mut new_plan = plan.clone();
- if !ordering_satisfy_requirement(
- plan.output_ordering(),
- parent_required,
- || plan.equivalence_properties(),
- || plan.ordering_equivalence_properties(),
- ) {
+ let new_plan = if !plan
Review Comment:
This is a really good example of how much nicer the new API is. 👨🍳 👌 --
really nice
##########
datafusion/physical-expr/src/sort_expr.rs:
##########
@@ -77,18 +77,26 @@ impl PhysicalSortExpr {
})
}
- /// Check whether sort expression satisfies [`PhysicalSortRequirement`].
- ///
- /// If sort options is Some in `PhysicalSortRequirement`, `expr`
- /// and `options` field are compared for equality.
- ///
- /// If sort options is None in `PhysicalSortRequirement`, only
- /// `expr` is compared for equality.
- pub fn satisfy(&self, requirement: &PhysicalSortRequirement) -> bool {
+ /// Checks whether this sort expression satisfies the given `requirement`.
+ /// If sort options are unspecified in `requirement`, only expressions are
+ /// compared for inequality.
+ pub fn satisfy(
+ &self,
+ requirement: &PhysicalSortRequirement,
+ schema: &Schema,
+ ) -> bool {
+ // If the column is not nullable, NULLS FIRST/LAST is not important.
Review Comment:
👍
##########
datafusion/physical-expr/src/sort_expr.rs:
##########
@@ -77,18 +77,26 @@ impl PhysicalSortExpr {
})
}
- /// Check whether sort expression satisfies [`PhysicalSortRequirement`].
- ///
- /// If sort options is Some in `PhysicalSortRequirement`, `expr`
- /// and `options` field are compared for equality.
- ///
- /// If sort options is None in `PhysicalSortRequirement`, only
- /// `expr` is compared for equality.
- pub fn satisfy(&self, requirement: &PhysicalSortRequirement) -> bool {
+ /// Checks whether this sort expression satisfies the given `requirement`.
+ /// If sort options are unspecified in `requirement`, only expressions are
+ /// compared for inequality.
+ pub fn satisfy(
+ &self,
+ requirement: &PhysicalSortRequirement,
+ schema: &Schema,
+ ) -> bool {
+ // If the column is not nullable, NULLS FIRST/LAST is not important.
Review Comment:
👍
--
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]
