MinMaxAggregateFn

Signed-off-by: Nicholas Gates <nick@nickgates.com>

Author: gatesn

encodings/runend/src/compute/mod.rs

@@ -7,6 +7,7 @@ mod fill_null;
 pub(crate) mod filter;
 mod is_constant;
 mod is_sorted;
+pub(crate) mod min_max;
 pub(crate) mod take;
 pub(crate) mod take_from;
 

encodings/runend/src/lib.rs

@@ -26,12 +26,22 @@ pub mod _benchmarking {
     use super::*;
 }
 
+use vortex_array::aggregate_fn::AggregateFnVTable;
+use vortex_array::aggregate_fn::fns::min_max::MinMax;
+use vortex_array::aggregate_fn::session::AggregateFnSessionExt;
 use vortex_array::session::ArraySessionExt;
 use vortex_session::VortexSession;
 
 /// Initialize run-end encoding in the given session.
 pub fn initialize(session: &mut VortexSession) {
     session.arrays().register(RunEnd::ID, RunEnd);
+
+    // Register the RunEnd-specific min/max aggregate kernel.
+    session.aggregate_fns().register_aggregate_kernel(
+        RunEnd::ID,
+        Some(MinMax.id()),
+        &compute::min_max::RunEndMinMaxKernel,
+    );
 }
 
 #[cfg(test)]

encodings/sequence/public-api.lock

@@ -158,4 +158,6 @@ pub vortex_sequence::SequenceArrayParts::nullability: vortex_array::dtype::nulla
 
 pub vortex_sequence::SequenceArrayParts::ptype: vortex_array::dtype::ptype::PType
 
+pub fn vortex_sequence::initialize(session: &mut vortex_session::VortexSession)
+
 pub fn vortex_sequence::sequence_encode(primitive_array: &vortex_array::arrays::primitive::array::PrimitiveArray) -> vortex_error::VortexResult<core::option::Option<vortex_array::array::ArrayRef>>

encodings/sequence/src/compute/mod.rs

@@ -6,6 +6,7 @@ pub(crate) mod compare;
 mod filter;
 mod is_sorted;
 mod list_contains;
+pub(crate) mod min_max;
 mod slice;
 mod take;
 

encodings/sequence/src/lib.rs

@@ -15,6 +15,23 @@ pub use array::Sequence;
 pub use array::SequenceArray;
 pub use array::SequenceArrayParts;
 pub use compress::sequence_encode;
+use vortex_array::aggregate_fn::AggregateFnVTable;
+use vortex_array::aggregate_fn::fns::min_max::MinMax;
+use vortex_array::aggregate_fn::session::AggregateFnSessionExt;
+use vortex_array::session::ArraySessionExt;
+use vortex_session::VortexSession;
+
+/// Initialize sequence encoding in the given session.
+pub fn initialize(session: &mut VortexSession) {
+    session.arrays().register(Sequence::ID, Sequence);
+
+    // Register the Sequence-specific min/max aggregate kernel.
+    session.aggregate_fns().register_aggregate_kernel(
+        Sequence::ID,
+        Some(MinMax.id()),
+        &compute::min_max::SequenceMinMaxKernel,
+    );
+}
 
 // TODO(joe): hook up to the compressor
 // TODO(joe): support comparisons with other operators

vortex-array/public-api.lock

@@ -102,6 +102,8 @@ pub fn vortex_array::aggregate_fn::fns::min_max::MinMaxResult::fmt(&self, f: &mu
 
 impl core::marker::StructuralPartialEq for vortex_array::aggregate_fn::fns::min_max::MinMaxResult
 
+pub fn vortex_array::aggregate_fn::fns::min_max::make_struct_dtype(element_dtype: &vortex_array::dtype::DType) -> vortex_array::dtype::DType
+
 pub fn vortex_array::aggregate_fn::fns::min_max::min_max(array: &vortex_array::ArrayRef, ctx: &mut vortex_array::ExecutionCtx) -> vortex_error::VortexResult<core::option::Option<vortex_array::aggregate_fn::fns::min_max::MinMaxResult>>
 
 pub mod vortex_array::aggregate_fn::fns::nan_count

vortex-array/src/aggregate_fn/fns/min_max/mod.rs

@@ -154,7 +154,8 @@ impl MinMaxPartial {
     }
 }
 
-fn make_struct_dtype(element_dtype: &DType) -> DType {
+/// Creates the struct dtype `{min: T, max: T}` (nullable) used for min/max aggregate results.
+pub fn make_struct_dtype(element_dtype: &DType) -> DType {
     DType::Struct(
         StructFields::new(
             NAMES.clone(),

vortex-array/src/aggregate_fn/session.rs

@@ -12,10 +12,13 @@ use vortex_utils::aliases::hash_map::HashMap;
 use crate::aggregate_fn::AggregateFnId;
 use crate::aggregate_fn::AggregateFnPluginRef;
 use crate::aggregate_fn::AggregateFnVTable;
+use crate::aggregate_fn::fns::min_max::MinMax;
 use crate::aggregate_fn::kernels::DynAggregateKernel;
 use crate::aggregate_fn::kernels::DynGroupedAggregateKernel;
 use crate::arrays::Chunked;
+use crate::arrays::Dict;
 use crate::arrays::chunked::compute::aggregate::ChunkedArrayAggregate;
+use crate::arrays::dict::compute::min_max::DictMinMaxKernel;
 use crate::vtable::ArrayId;
 
 /// Registry of aggregate function vtables.
@@ -42,6 +45,7 @@ impl Default for AggregateFnSession {
 
         // Register the built-in aggregate kernels.
         this.register_aggregate_kernel(Chunked::ID, None, &ChunkedArrayAggregate);
+        this.register_aggregate_kernel(Dict::ID, Some(MinMax.id()), &DictMinMaxKernel);
 
         this
     }

vortex-array/src/arrays/dict/compute/mod.rs

@@ -8,6 +8,7 @@ mod is_constant;
 mod is_sorted;
 mod like;
 mod mask;
+pub(crate) mod min_max;
 pub(crate) mod rules;
 mod slice;
 

vortex-file/src/lib.rs

@@ -120,7 +120,6 @@ use vortex_fastlanes::FoR;
 use vortex_fastlanes::RLE;
 use vortex_fsst::FSST;
 use vortex_pco::Pco;
-use vortex_sequence::Sequence;
 use vortex_session::VortexSession;
 use vortex_sparse::Sparse;
 use vortex_zigzag::ZigZag;
@@ -175,7 +174,6 @@ pub fn register_default_encodings(session: &mut VortexSession) {
         arrays.register(FoR::ID, FoR);
         arrays.register(Pco::ID, Pco);
         arrays.register(RLE::ID, RLE);
-        arrays.register(Sequence::ID, Sequence);
         arrays.register(Sparse::ID, Sparse);
         arrays.register(ZigZag::ID, ZigZag);
         #[cfg(feature = "zstd")]
@@ -187,5 +185,6 @@ pub fn register_default_encodings(session: &mut VortexSession) {
     // Eventually all encodings crates should expose an initialize function. For now it's only
     // a few of them.
     vortex_alp::initialize(session);
-    vortex_runend::initialize(session)
+    vortex_runend::initialize(session);
+    vortex_sequence::initialize(session);
 }