add count distinct group benchmarks

coderfender · coderfender · commit 36f232c6496e · 2026-04-19T09:20:49.000-07:00
diff --git a/datafusion/functions-aggregate-common/src/aggregate/count_distinct.rs b/datafusion/functions-aggregate-common/src/aggregate/count_distinct.rs
@@ -17,14 +17,19 @@
 
 mod bytes;
 mod dict;
+mod groups;
 mod native;
 
 pub use bytes::BytesDistinctCountAccumulator;
 pub use bytes::BytesViewDistinctCountAccumulator;
 pub use dict::DictionaryCountAccumulator;
+<<<<<<< HEAD
 pub use native::Bitmap65536DistinctCountAccumulator;
 pub use native::Bitmap65536DistinctCountAccumulatorI16;
 pub use native::BoolArray256DistinctCountAccumulator;
 pub use native::BoolArray256DistinctCountAccumulatorI8;
+=======
+pub use groups::PrimitiveDistinctCountGroupsAccumulator;
+>>>>>>> f982d8de2 (add count distinct group benchmarks)
 pub use native::FloatDistinctCountAccumulator;
 pub use native::PrimitiveDistinctCountAccumulator;
diff --git a/datafusion/functions-aggregate-common/src/aggregate/count_distinct/groups.rs b/datafusion/functions-aggregate-common/src/aggregate/count_distinct/groups.rs
@@ -0,0 +1,182 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements.  See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership.  The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License.  You may obtain a copy of the License at
+//
+//   http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied.  See the License for the
+// specific language governing permissions and limitations
+// under the License.
+
+use arrow::array::{
+    ArrayRef, AsArray, BooleanArray, Int64Array, ListArray, PrimitiveArray,
+};
+use arrow::buffer::OffsetBuffer;
+use arrow::datatypes::{ArrowPrimitiveType, Field};
+use datafusion_common::HashSet;
+use datafusion_common::hash_utils::RandomState;
+use datafusion_expr_common::groups_accumulator::{EmitTo, GroupsAccumulator};
+use std::hash::Hash;
+use std::mem::size_of;
+use std::sync::Arc;
+
+use crate::aggregate::groups_accumulator::accumulate::accumulate;
+
+pub struct PrimitiveDistinctCountGroupsAccumulator<T: ArrowPrimitiveType>
+where
+    T::Native: Eq + Hash,
+{
+    seen: HashSet<(usize, T::Native), RandomState>,
+    num_groups: usize,
+}
+
+impl<T: ArrowPrimitiveType> PrimitiveDistinctCountGroupsAccumulator<T>
+where
+    T::Native: Eq + Hash,
+{
+    pub fn new() -> Self {
+        Self {
+            seen: HashSet::default(),
+            num_groups: 0,
+        }
+    }
+}
+
+impl<T: ArrowPrimitiveType> Default for PrimitiveDistinctCountGroupsAccumulator<T>
+where
+    T::Native: Eq + Hash,
+{
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+impl<T: ArrowPrimitiveType + Send + std::fmt::Debug> GroupsAccumulator
+    for PrimitiveDistinctCountGroupsAccumulator<T>
+where
+    T::Native: Eq + Hash,
+{
+    fn update_batch(
+        &mut self,
+        values: &[ArrayRef],
+        group_indices: &[usize],
+        opt_filter: Option<&BooleanArray>,
+        total_num_groups: usize,
+    ) -> datafusion_common::Result<()> {
+        debug_assert_eq!(values.len(), 1);
+        self.num_groups = self.num_groups.max(total_num_groups);
+        let arr = values[0].as_primitive::<T>();
+        accumulate(group_indices, arr, opt_filter, |group_idx, value| {
+            self.seen.insert((group_idx, value));
+        });
+        Ok(())
+    }
+
+    fn evaluate(&mut self, emit_to: EmitTo) -> datafusion_common::Result<ArrayRef> {
+        let num_emitted = match emit_to {
+            EmitTo::All => self.num_groups,
+            EmitTo::First(n) => n,
+        };
+
+        let mut counts = vec![0i64; num_emitted];
+
+        if matches!(emit_to, EmitTo::All) {
+            for &(group_idx, _) in self.seen.iter() {
+                counts[group_idx] += 1;
+            }
+            self.seen.clear();
+            self.num_groups = 0;
+        } else {
+            let mut remaining = HashSet::default();
+            for (group_idx, value) in self.seen.drain() {
+                if group_idx < num_emitted {
+                    counts[group_idx] += 1;
+                } else {
+                    remaining.insert((group_idx - num_emitted, value));
+                }
+            }
+            self.seen = remaining;
+            self.num_groups = self.num_groups.saturating_sub(num_emitted);
+        }
+
+        Ok(Arc::new(Int64Array::from(counts)))
+    }
+
+    fn state(&mut self, emit_to: EmitTo) -> datafusion_common::Result<Vec<ArrayRef>> {
+        let num_emitted = match emit_to {
+            EmitTo::All => self.num_groups,
+            EmitTo::First(n) => n,
+        };
+
+        let mut group_values: Vec<Vec<T::Native>> = vec![Vec::new(); num_emitted];
+
+        if matches!(emit_to, EmitTo::All) {
+            for (group_idx, value) in self.seen.drain() {
+                group_values[group_idx].push(value);
+            }
+            self.num_groups = 0;
+        } else {
+            let mut remaining = HashSet::default();
+            for (group_idx, value) in self.seen.drain() {
+                if group_idx < num_emitted {
+                    group_values[group_idx].push(value);
+                } else {
+                    remaining.insert((group_idx - num_emitted, value));
+                }
+            }
+            self.seen = remaining;
+            self.num_groups = self.num_groups.saturating_sub(num_emitted);
+        }
+
+        let mut offsets = vec![0i32];
+        let mut all_values = Vec::new();
+        for values in &group_values {
+            all_values.extend(values.iter().copied());
+            offsets.push(all_values.len() as i32);
+        }
+
+        let values_array = Arc::new(PrimitiveArray::<T>::from_iter_values(all_values));
+        let list_array = ListArray::new(
+            Arc::new(Field::new_list_field(T::DATA_TYPE, true)),
+            OffsetBuffer::new(offsets.into()),
+            values_array,
+            None,
+        );
+
+        Ok(vec![Arc::new(list_array)])
+    }
+
+    fn merge_batch(
+        &mut self,
+        values: &[ArrayRef],
+        group_indices: &[usize],
+        _opt_filter: Option<&BooleanArray>,
+        total_num_groups: usize,
+    ) -> datafusion_common::Result<()> {
+        debug_assert_eq!(values.len(), 1);
+        self.num_groups = self.num_groups.max(total_num_groups);
+        let list_array = values[0].as_list::<i32>();
+
+        for (row_idx, group_idx) in group_indices.iter().enumerate() {
+            let inner = list_array.value(row_idx);
+            let inner_arr = inner.as_primitive::<T>();
+            for value in inner_arr.values().iter() {
+                self.seen.insert((*group_idx, *value));
+            }
+        }
+
+        Ok(())
+    }
+
+    fn size(&self) -> usize {
+        size_of::<Self>()
+            + self.seen.capacity() * (size_of::<(usize, T::Native)>() + size_of::<u64>())
+    }
+}
diff --git a/datafusion/functions-aggregate/benches/count_distinct.rs b/datafusion/functions-aggregate/benches/count_distinct.rs
@@ -18,7 +18,7 @@
 use std::sync::Arc;
 
 use arrow::array::{
-    Array, ArrayRef, Int8Array, Int16Array, Int64Array, UInt8Array, UInt16Array,
+    ArrayRef, Int8Array, Int16Array, Int64Array, UInt8Array, UInt16Array,
 };
 use arrow::datatypes::{DataType, Field, Schema};
 use criterion::{Criterion, criterion_group, criterion_main};
@@ -87,6 +87,13 @@ fn create_i16_array(n_distinct: usize) -> Int16Array {
         .collect()
 }
 
+fn create_group_indices(num_groups: usize) -> Vec<usize> {
+    let mut rng = StdRng::seed_from_u64(42);
+    (0..BATCH_SIZE)
+        .map(|_| rng.random_range(0..num_groups))
+        .collect()
+}
+
 fn prepare_args(data_type: DataType) -> (Arc<Schema>, AccumulatorArgs<'static>) {
     let schema = Arc::new(Schema::new(vec![Field::new("f", data_type, true)]));
     let schema_leaked: &'static Schema = Box::leak(Box::new((*schema).clone()));
@@ -174,103 +181,62 @@ fn count_distinct_benchmark(c: &mut Criterion) {
     });
 }
 
-/// Create group indices with uniform distribution
-fn create_uniform_groups(num_groups: usize) -> Vec<usize> {
-    let mut rng = StdRng::seed_from_u64(42);
-    (0..BATCH_SIZE)
-        .map(|_| rng.random_range(0..num_groups))
-        .collect()
-}
-
-/// Create group indices with skewed distribution (80% in 20% of groups)
-fn create_skewed_groups(num_groups: usize) -> Vec<usize> {
-    let mut rng = StdRng::seed_from_u64(42);
-    let hot_groups = (num_groups / 5).max(1);
-    (0..BATCH_SIZE)
-        .map(|_| {
-            if rng.random_range(0..100) < 80 {
-                rng.random_range(0..hot_groups)
-            } else {
-                rng.random_range(0..num_groups)
-            }
-        })
-        .collect()
-}
-
 fn count_distinct_groups_benchmark(c: &mut Criterion) {
     let count_fn = Count::new();
 
-    // bench different scenarios
-    let scenarios = [
-        // (name, num_groups, distinct_pct, group_fn)
-        ("sparse_uniform", 10, 80, "uniform"),
-        ("moderate_uniform", 100, 80, "uniform"),
-        ("dense_uniform", 1000, 80, "uniform"),
-        ("sparse_skewed", 10, 80, "skewed"),
-        ("dense_skewed", 1000, 80, "skewed"),
-        ("sparse_high_cardinality", 10, 99, "uniform"),
-        ("dense_low_cardinality", 1000, 20, "uniform"),
-    ];
-
-    for (name, num_groups, distinct_pct, group_type) in scenarios {
-        let n_distinct = BATCH_SIZE * distinct_pct / 100;
+    for num_groups in [10, 100, 1000] {
+        let n_distinct = BATCH_SIZE * 80 / 100;
         let values = Arc::new(create_i64_array(n_distinct)) as ArrayRef;
-        let group_indices = if group_type == "uniform" {
-            create_uniform_groups(num_groups)
-        } else {
-            create_skewed_groups(num_groups)
-        };
+        let group_indices = create_group_indices(num_groups);
 
         let (_schema, args) = prepare_args(DataType::Int64);
 
         if count_fn.groups_accumulator_supported(args.clone()) {
-            c.bench_function(&format!("count_distinct_groups {name}"), |b| {
-                b.iter(|| {
-                    let mut acc =
-                        count_fn.create_groups_accumulator(args.clone()).unwrap();
-                    acc.update_batch(
-                        std::slice::from_ref(&values),
-                        &group_indices,
-                        None,
-                        num_groups,
-                    )
-                    .unwrap();
-                    acc.evaluate(EmitTo::All).unwrap()
-                })
-            });
+            c.bench_function(
+                &format!("count_distinct_groups i64 {num_groups} groups"),
+                |b| {
+                    b.iter(|| {
+                        let (_schema, args) = prepare_args(DataType::Int64);
+                        let mut acc = count_fn.create_groups_accumulator(args).unwrap();
+                        acc.update_batch(
+                            std::slice::from_ref(&values),
+                            &group_indices,
+                            None,
+                            num_groups,
+                        )
+                        .unwrap();
+                        acc.evaluate(EmitTo::All).unwrap()
+                    })
+                },
+            );
         } else {
-            let arr = values.as_any().downcast_ref::<Int64Array>().unwrap();
-            let mut group_rows: Vec<Vec<i64>> = vec![Vec::new(); num_groups];
-            for (idx, &group_idx) in group_indices.iter().enumerate() {
-                if arr.is_valid(idx) {
-                    group_rows[group_idx].push(arr.value(idx));
-                }
-            }
-            let group_arrays: Vec<ArrayRef> = group_rows
-                .iter()
-                .map(|rows| Arc::new(Int64Array::from(rows.clone())) as ArrayRef)
-                .collect();
-
-            c.bench_function(&format!("count_distinct_groups {name}"), |b| {
-                b.iter(|| {
-                    let mut accumulators: Vec<_> = (0..num_groups)
-                        .map(|_| prepare_accumulator(DataType::Int64))
-                        .collect();
-
-                    for (group_idx, batch) in group_arrays.iter().enumerate() {
-                        if !batch.is_empty() {
-                            accumulators[group_idx]
-                                .update_batch(std::slice::from_ref(batch))
-                                .unwrap();
+            c.bench_function(
+                &format!("count_distinct_groups i64 {num_groups} groups"),
+                |b| {
+                    b.iter(|| {
+                        let mut accumulators: Vec<_> = (0..num_groups)
+                            .map(|_| prepare_accumulator(DataType::Int64))
+                            .collect();
+
+                        let arr = values.as_any().downcast_ref::<Int64Array>().unwrap();
+                        for (idx, group_idx) in group_indices.iter().enumerate() {
+                            if let Some(val) = arr.value(idx).into() {
+                                let single_val =
+                                    Arc::new(Int64Array::from(vec![Some(val)]))
+                                        as ArrayRef;
+                                accumulators[*group_idx]
+                                    .update_batch(std::slice::from_ref(&single_val))
+                                    .unwrap();
+                            }
                         }
-                    }
 
-                    let _results: Vec<_> = accumulators
-                        .iter_mut()
-                        .map(|acc| acc.evaluate().unwrap())
-                        .collect();
-                })
-            });
+                        let _results: Vec<_> = accumulators
+                            .iter_mut()
+                            .map(|acc| acc.evaluate().unwrap())
+                            .collect();
+                    })
+                },
+            );
         }
     }
 }
diff --git a/datafusion/functions-aggregate/src/count.rs b/datafusion/functions-aggregate/src/count.rs
