Skip to content

fix: UDF, UDAF, UDWF with_alias(..) should wrap the inner function fully #12098

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 5 commits into from
Aug 22, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
61 changes: 61 additions & 0 deletions datafusion/expr/src/udaf.rs
Original file line number Diff line number Diff line change
Expand Up @@ -337,6 +337,9 @@ where
/// let expr = geometric_mean.call(vec![col("a")]);
/// ```
pub trait AggregateUDFImpl: Debug + Send + Sync {
// Note: When adding any methods (with default implementations), remember to add them also
// into the AliasedAggregateUDFImpl below!

/// Returns this object as an [`Any`] trait object
fn as_any(&self) -> &dyn Any;

Expand Down Expand Up @@ -635,6 +638,60 @@ impl AggregateUDFImpl for AliasedAggregateUDFImpl {
&self.aliases
}

fn state_fields(&self, args: StateFieldsArgs) -> Result<Vec<Field>> {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

👍 nice - to avoid similar bugs in the future maybe we should also add a comment to AggregateUDFImpl that tries to remind people to add an implementation here.

It would be cool if we could tell Rust not to create a default implementation for a particular impl so the compiler could do the check for us, but I don't know how to express that in Rust

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Added notes in 175b8d2. And yep - it would be nice to have a way of enforcing this, but I also couldn't figure out any way to do that :/

self.inner.state_fields(args)
}

fn groups_accumulator_supported(&self, args: AccumulatorArgs) -> bool {
self.inner.groups_accumulator_supported(args)
}

fn create_groups_accumulator(
&self,
args: AccumulatorArgs,
) -> Result<Box<dyn GroupsAccumulator>> {
self.inner.create_groups_accumulator(args)
}

fn create_sliding_accumulator(
&self,
args: AccumulatorArgs,
) -> Result<Box<dyn Accumulator>> {
self.inner.accumulator(args)
}

fn with_beneficial_ordering(
self: Arc<Self>,
beneficial_ordering: bool,
) -> Result<Option<Arc<dyn AggregateUDFImpl>>> {
Arc::clone(&self.inner)
.with_beneficial_ordering(beneficial_ordering)
.map(|udf| {
udf.map(|udf| {
Arc::new(AliasedAggregateUDFImpl {
inner: udf,
aliases: self.aliases.clone(),
}) as Arc<dyn AggregateUDFImpl>
})
})
}

fn order_sensitivity(&self) -> AggregateOrderSensitivity {
self.inner.order_sensitivity()
}

fn simplify(&self) -> Option<AggregateFunctionSimplification> {
self.inner.simplify()
}

fn reverse_expr(&self) -> ReversedUDAF {
self.inner.reverse_expr()
}

fn coerce_types(&self, arg_types: &[DataType]) -> Result<Vec<DataType>> {
self.inner.coerce_types(arg_types)
}

fn equals(&self, other: &dyn AggregateUDFImpl) -> bool {
if let Some(other) = other.as_any().downcast_ref::<AliasedAggregateUDFImpl>() {
self.inner.equals(other.inner.as_ref()) && self.aliases == other.aliases
Expand All @@ -649,6 +706,10 @@ impl AggregateUDFImpl for AliasedAggregateUDFImpl {
self.aliases.hash(hasher);
hasher.finish()
}

fn is_descending(&self) -> Option<bool> {
self.inner.is_descending()
}
}

/// Implementation of [`AggregateUDFImpl`] that wraps the function style pointers
Expand Down
60 changes: 58 additions & 2 deletions datafusion/expr/src/udf.rs
Original file line number Diff line number Diff line change
Expand Up @@ -346,6 +346,9 @@ where
/// let expr = add_one.call(vec![col("a")]);
/// ```
pub trait ScalarUDFImpl: Debug + Send + Sync {
// Note: When adding any methods (with default implementations), remember to add them also
// into the AliasedScalarUDFImpl below!

/// Returns this object as an [`Any`] trait object
fn as_any(&self) -> &dyn Any;

Expand Down Expand Up @@ -632,6 +635,14 @@ impl ScalarUDFImpl for AliasedScalarUDFImpl {
self.inner.name()
}

fn display_name(&self, args: &[Expr]) -> Result<String> {
self.inner.display_name(args)
}

fn schema_name(&self, args: &[Expr]) -> Result<String> {
self.inner.schema_name(args)
}

fn signature(&self) -> &Signature {
self.inner.signature()
}
Expand All @@ -640,12 +651,57 @@ impl ScalarUDFImpl for AliasedScalarUDFImpl {
self.inner.return_type(arg_types)
}

fn aliases(&self) -> &[String] {
&self.aliases
}

fn return_type_from_exprs(
&self,
args: &[Expr],
schema: &dyn ExprSchema,
arg_types: &[DataType],
) -> Result<DataType> {
self.inner.return_type_from_exprs(args, schema, arg_types)
}

fn invoke(&self, args: &[ColumnarValue]) -> Result<ColumnarValue> {
self.inner.invoke(args)
}

fn aliases(&self) -> &[String] {
&self.aliases
fn invoke_no_args(&self, number_rows: usize) -> Result<ColumnarValue> {
self.inner.invoke_no_args(number_rows)
}

fn simplify(
&self,
args: Vec<Expr>,
info: &dyn SimplifyInfo,
) -> Result<ExprSimplifyResult> {
self.inner.simplify(args, info)
}

fn short_circuits(&self) -> bool {
self.inner.short_circuits()
}

fn evaluate_bounds(&self, input: &[&Interval]) -> Result<Interval> {
self.inner.evaluate_bounds(input)
}

fn propagate_constraints(
&self,
interval: &Interval,
inputs: &[&Interval],
) -> Result<Option<Vec<Interval>>> {
self.inner.propagate_constraints(interval, inputs)
}

fn output_ordering(&self, inputs: &[ExprProperties]) -> Result<SortProperties> {
self.inner.output_ordering(inputs)
}

fn coerce_types(&self, arg_types: &[DataType]) -> Result<Vec<DataType>> {
self.inner.coerce_types(arg_types)
}

fn equals(&self, other: &dyn ScalarUDFImpl) -> bool {
Expand Down
19 changes: 19 additions & 0 deletions datafusion/expr/src/udwf.rs
Original file line number Diff line number Diff line change
Expand Up @@ -266,6 +266,9 @@ where
/// .unwrap();
/// ```
pub trait WindowUDFImpl: Debug + Send + Sync {
// Note: When adding any methods (with default implementations), remember to add them also
// into the AliasedWindowUDFImpl below!

/// Returns this object as an [`Any`] trait object
fn as_any(&self) -> &dyn Any;

Expand Down Expand Up @@ -428,6 +431,10 @@ impl WindowUDFImpl for AliasedWindowUDFImpl {
&self.aliases
}

fn simplify(&self) -> Option<WindowFunctionSimplification> {
self.inner.simplify()
}

fn equals(&self, other: &dyn WindowUDFImpl) -> bool {
if let Some(other) = other.as_any().downcast_ref::<AliasedWindowUDFImpl>() {
self.inner.equals(other.inner.as_ref()) && self.aliases == other.aliases
Expand All @@ -442,6 +449,18 @@ impl WindowUDFImpl for AliasedWindowUDFImpl {
self.aliases.hash(hasher);
hasher.finish()
}

fn nullable(&self) -> bool {
self.inner.nullable()
}

fn sort_options(&self) -> Option<SortOptions> {
self.inner.sort_options()
}

fn coerce_types(&self, arg_types: &[DataType]) -> Result<Vec<DataType>> {
self.inner.coerce_types(arg_types)
}
}

/// Implementation of [`WindowUDFImpl`] that wraps the function style pointers
Expand Down