[GitHub] [arrow] jorgecarleitao commented on a change in pull request #8222: ARROW-10043: [Rust][DataFusion] Implement COUNT(DISTINCT col)

[GitBox <gi...@apache.org>]

jorgecarleitao commented on a change in pull request #8222:
URL: https://github.com/apache/arrow/pull/8222#discussion_r491530453



##########
File path: rust/datafusion/src/physical_plan/distinct_expressions.rs
##########
@@ -0,0 +1,303 @@
+// 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.
+
+//! Implementations for DISTINCT expressions, e.g. `COUNT(DISTINCT c)`
+
+use std::cell::RefCell;
+use std::fmt::Debug;
+use std::hash::Hash;
+use std::rc::Rc;
+use std::sync::Arc;
+
+use arrow::array::ArrayRef;
+use arrow::array::{
+    Int16Array, Int32Array, Int64Array, Int8Array, PrimitiveArrayOps, UInt16Array,
+    UInt32Array, UInt64Array, UInt8Array,
+};
+use arrow::datatypes::{DataType, Schema};
+use arrow::record_batch::RecordBatch;
+
+use fnv::FnvHashSet;
+
+use crate::error::{ExecutionError, Result};
+use crate::logical_plan::ScalarValue;
+use crate::physical_plan::expressions::Column;
+use crate::physical_plan::hash_aggregate::AggregateMode;
+use crate::physical_plan::{Accumulator, AggregateExpr, PhysicalExpr};
+
+/// Enumeration of types that can be accumulated into a distinct set of values.
+#[derive(Debug, PartialEq, Eq, Hash, Clone)]
+enum DistinctScalarValue {

Review comment:
       Wouldn't it be possible to use `ScalarValue`, instead of declaring a new enum?

##########
File path: rust/datafusion/src/physical_plan/hash_aggregate.rs
##########
@@ -48,9 +48,12 @@ use fnv::FnvHashMap;
 /// Hash aggregate modes
 #[derive(Debug, Copy, Clone)]
 pub enum AggregateMode {
+    /// Aggregate mode without any partial/intermediate aggregation
+    NoPartial,

Review comment:
       A naming idea: `SinglePartition`

##########
File path: rust/datafusion/src/physical_plan/distinct_expressions.rs
##########
@@ -0,0 +1,303 @@
+// 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.
+
+//! Implementations for DISTINCT expressions, e.g. `COUNT(DISTINCT c)`
+
+use std::cell::RefCell;
+use std::fmt::Debug;
+use std::hash::Hash;
+use std::rc::Rc;
+use std::sync::Arc;
+
+use arrow::array::ArrayRef;
+use arrow::array::{
+    Int16Array, Int32Array, Int64Array, Int8Array, PrimitiveArrayOps, UInt16Array,
+    UInt32Array, UInt64Array, UInt8Array,
+};
+use arrow::datatypes::{DataType, Schema};
+use arrow::record_batch::RecordBatch;
+
+use fnv::FnvHashSet;
+
+use crate::error::{ExecutionError, Result};
+use crate::logical_plan::ScalarValue;
+use crate::physical_plan::expressions::Column;
+use crate::physical_plan::hash_aggregate::AggregateMode;
+use crate::physical_plan::{Accumulator, AggregateExpr, PhysicalExpr};
+
+/// Enumeration of types that can be accumulated into a distinct set of values.
+#[derive(Debug, PartialEq, Eq, Hash, Clone)]
+enum DistinctScalarValue {
+    Int8(i8),
+    Int16(i16),
+    Int32(i32),
+    Int64(i64),
+    UInt8(u8),
+    UInt16(u16),
+    UInt32(u32),
+    UInt64(u64),
+}
+
+/// For a given expression, maps its Arrow DataType into a LargeList of the
+/// same DataType.
+fn list_of(expr: Arc<dyn PhysicalExpr>, input_schema: &Schema) -> Result<DataType> {
+    let value_data_type = expr.data_type(input_schema)?;
+
+    match value_data_type {
+        DataType::Int8
+        | DataType::Int16
+        | DataType::Int32
+        | DataType::Int64
+        | DataType::UInt8
+        | DataType::UInt16
+        | DataType::UInt32
+        | DataType::UInt64 => Ok(DataType::LargeList(Box::new(value_data_type))),
+        _ => Err(ExecutionError::NotImplemented(
+            "Unsupported column data type for DISTINCT".to_string(),
+        )),
+    }
+}
+
+fn accumulate_scalar(
+    accum: &mut FnvHashSet<DistinctScalarValue>,
+    value: Option<ScalarValue>,
+) -> Result<()> {
+    let accum_value = match value {
+        Some(ScalarValue::Int8(v)) => Some(DistinctScalarValue::Int8(v)),
+        Some(ScalarValue::Int16(v)) => Some(DistinctScalarValue::Int16(v)),
+        Some(ScalarValue::Int32(v)) => Some(DistinctScalarValue::Int32(v)),
+        Some(ScalarValue::Int64(v)) => Some(DistinctScalarValue::Int64(v)),
+        Some(ScalarValue::UInt8(v)) => Some(DistinctScalarValue::UInt8(v)),
+        Some(ScalarValue::UInt16(v)) => Some(DistinctScalarValue::UInt16(v)),
+        Some(ScalarValue::UInt32(v)) => Some(DistinctScalarValue::UInt32(v)),
+        Some(ScalarValue::UInt64(v)) => Some(DistinctScalarValue::UInt64(v)),
+        Some(ScalarValue::Null) => None,
+        _ => {
+            return Err(ExecutionError::NotImplemented(
+                "Unsupported scalar value for DISTINCT accumulator".to_string(),
+            ))
+        }
+    };
+
+    match accum_value {
+        Some(v) => {
+            accum.insert(v);
+        }
+        None => {}
+    }
+
+    Ok(())
+}
+
+macro_rules! accum_batch {
+    ($ARRAY_TY:ident, $DISTINCT_SCALAR_TY: path, $ARRAY: expr, $ACCUM: expr) => {{
+        let array = $ARRAY.as_any().downcast_ref::<$ARRAY_TY>().ok_or_else(|| {
+            ExecutionError::ExecutionError("Error downcasting array".to_string())
+        })?;
+
+        for i in 0..array.len() {
+            $ACCUM.insert($DISTINCT_SCALAR_TY(array.value(i)));
+        }
+
+        Ok(())
+    }};
+}
+
+#[derive(Debug)]
+struct DistinctValuesAccumulator {
+    values: FnvHashSet<DistinctScalarValue>,
+}
+
+impl Accumulator for DistinctValuesAccumulator {
+    fn accumulate_scalar(&mut self, value: Option<ScalarValue>) -> Result<()> {
+        accumulate_scalar(&mut self.values, value)
+    }
+
+    fn accumulate_batch(&mut self, _array: &ArrayRef) -> Result<()> {
+        Err(ExecutionError::NotImplemented(
+            "Aggregates with DISTINCT not supported without a GROUP BY".to_string(),
+        ))
+    }
+
+    fn get_value(&self) -> Result<Option<ScalarValue>> {
+        let value_out = self
+            .values
+            .iter()
+            .map(|accumulated_value| match accumulated_value {
+                DistinctScalarValue::Int8(v) => Ok(ScalarValue::Int8(*v)),
+                DistinctScalarValue::Int16(v) => Ok(ScalarValue::Int16(*v)),
+                DistinctScalarValue::Int32(v) => Ok(ScalarValue::Int32(*v)),
+                DistinctScalarValue::Int64(v) => Ok(ScalarValue::Int64(*v)),
+                DistinctScalarValue::UInt8(v) => Ok(ScalarValue::UInt8(*v)),
+                DistinctScalarValue::UInt16(v) => Ok(ScalarValue::UInt16(*v)),
+                DistinctScalarValue::UInt32(v) => Ok(ScalarValue::UInt32(*v)),
+                DistinctScalarValue::UInt64(v) => Ok(ScalarValue::UInt64(*v)),
+            })
+            .collect::<Result<Vec<ScalarValue>>>()?;
+
+        Ok(Some(ScalarValue::Struct(value_out)))
+    }
+}
+
+/// Create a distint count expression.
+pub fn count(mode: AggregateMode, expr: Arc<dyn PhysicalExpr>) -> Arc<dyn AggregateExpr> {

Review comment:
       Note that this will not apply any coercion to the arguments. We are trying to make `aggregates::create_aggregate_expr` be the entry point to create the physical expressions, to guarantee that the signature is correct and coercion rules apply.
   
   However, I see that this depends on the `AggregateMode`, which the current `create_aggregate_expr` does not support. #8172 addresses exactly this. :)

##########
File path: rust/datafusion/src/physical_plan/distinct_expressions.rs
##########
@@ -0,0 +1,303 @@
+// 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.
+
+//! Implementations for DISTINCT expressions, e.g. `COUNT(DISTINCT c)`
+
+use std::cell::RefCell;
+use std::fmt::Debug;
+use std::hash::Hash;
+use std::rc::Rc;
+use std::sync::Arc;
+
+use arrow::array::ArrayRef;
+use arrow::array::{
+    Int16Array, Int32Array, Int64Array, Int8Array, PrimitiveArrayOps, UInt16Array,
+    UInt32Array, UInt64Array, UInt8Array,
+};
+use arrow::datatypes::{DataType, Schema};
+use arrow::record_batch::RecordBatch;
+
+use fnv::FnvHashSet;
+
+use crate::error::{ExecutionError, Result};
+use crate::logical_plan::ScalarValue;
+use crate::physical_plan::expressions::Column;
+use crate::physical_plan::hash_aggregate::AggregateMode;
+use crate::physical_plan::{Accumulator, AggregateExpr, PhysicalExpr};
+
+/// Enumeration of types that can be accumulated into a distinct set of values.
+#[derive(Debug, PartialEq, Eq, Hash, Clone)]
+enum DistinctScalarValue {
+    Int8(i8),
+    Int16(i16),
+    Int32(i32),
+    Int64(i64),
+    UInt8(u8),
+    UInt16(u16),
+    UInt32(u32),
+    UInt64(u64),
+}
+
+/// For a given expression, maps its Arrow DataType into a LargeList of the
+/// same DataType.
+fn list_of(expr: Arc<dyn PhysicalExpr>, input_schema: &Schema) -> Result<DataType> {
+    let value_data_type = expr.data_type(input_schema)?;
+
+    match value_data_type {
+        DataType::Int8
+        | DataType::Int16
+        | DataType::Int32
+        | DataType::Int64
+        | DataType::UInt8
+        | DataType::UInt16
+        | DataType::UInt32
+        | DataType::UInt64 => Ok(DataType::LargeList(Box::new(value_data_type))),
+        _ => Err(ExecutionError::NotImplemented(
+            "Unsupported column data type for DISTINCT".to_string(),
+        )),
+    }
+}
+
+fn accumulate_scalar(
+    accum: &mut FnvHashSet<DistinctScalarValue>,
+    value: Option<ScalarValue>,
+) -> Result<()> {
+    let accum_value = match value {
+        Some(ScalarValue::Int8(v)) => Some(DistinctScalarValue::Int8(v)),
+        Some(ScalarValue::Int16(v)) => Some(DistinctScalarValue::Int16(v)),
+        Some(ScalarValue::Int32(v)) => Some(DistinctScalarValue::Int32(v)),
+        Some(ScalarValue::Int64(v)) => Some(DistinctScalarValue::Int64(v)),
+        Some(ScalarValue::UInt8(v)) => Some(DistinctScalarValue::UInt8(v)),
+        Some(ScalarValue::UInt16(v)) => Some(DistinctScalarValue::UInt16(v)),
+        Some(ScalarValue::UInt32(v)) => Some(DistinctScalarValue::UInt32(v)),
+        Some(ScalarValue::UInt64(v)) => Some(DistinctScalarValue::UInt64(v)),
+        Some(ScalarValue::Null) => None,
+        _ => {
+            return Err(ExecutionError::NotImplemented(
+                "Unsupported scalar value for DISTINCT accumulator".to_string(),
+            ))
+        }
+    };
+
+    match accum_value {
+        Some(v) => {
+            accum.insert(v);
+        }
+        None => {}
+    }
+
+    Ok(())
+}
+
+macro_rules! accum_batch {
+    ($ARRAY_TY:ident, $DISTINCT_SCALAR_TY: path, $ARRAY: expr, $ACCUM: expr) => {{
+        let array = $ARRAY.as_any().downcast_ref::<$ARRAY_TY>().ok_or_else(|| {
+            ExecutionError::ExecutionError("Error downcasting array".to_string())
+        })?;
+
+        for i in 0..array.len() {
+            $ACCUM.insert($DISTINCT_SCALAR_TY(array.value(i)));
+        }
+
+        Ok(())
+    }};
+}
+
+#[derive(Debug)]
+struct DistinctValuesAccumulator {
+    values: FnvHashSet<DistinctScalarValue>,
+}
+
+impl Accumulator for DistinctValuesAccumulator {
+    fn accumulate_scalar(&mut self, value: Option<ScalarValue>) -> Result<()> {
+        accumulate_scalar(&mut self.values, value)
+    }
+
+    fn accumulate_batch(&mut self, _array: &ArrayRef) -> Result<()> {
+        Err(ExecutionError::NotImplemented(
+            "Aggregates with DISTINCT not supported without a GROUP BY".to_string(),
+        ))
+    }
+
+    fn get_value(&self) -> Result<Option<ScalarValue>> {
+        let value_out = self
+            .values
+            .iter()
+            .map(|accumulated_value| match accumulated_value {
+                DistinctScalarValue::Int8(v) => Ok(ScalarValue::Int8(*v)),

Review comment:
       ... and back: `DistinctScalarValue -> ScalarValue`

##########
File path: rust/datafusion/src/physical_plan/distinct_expressions.rs
##########
@@ -0,0 +1,303 @@
+// 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.
+
+//! Implementations for DISTINCT expressions, e.g. `COUNT(DISTINCT c)`
+
+use std::cell::RefCell;
+use std::fmt::Debug;
+use std::hash::Hash;
+use std::rc::Rc;
+use std::sync::Arc;
+
+use arrow::array::ArrayRef;
+use arrow::array::{
+    Int16Array, Int32Array, Int64Array, Int8Array, PrimitiveArrayOps, UInt16Array,
+    UInt32Array, UInt64Array, UInt8Array,
+};
+use arrow::datatypes::{DataType, Schema};
+use arrow::record_batch::RecordBatch;
+
+use fnv::FnvHashSet;
+
+use crate::error::{ExecutionError, Result};
+use crate::logical_plan::ScalarValue;
+use crate::physical_plan::expressions::Column;
+use crate::physical_plan::hash_aggregate::AggregateMode;
+use crate::physical_plan::{Accumulator, AggregateExpr, PhysicalExpr};
+
+/// Enumeration of types that can be accumulated into a distinct set of values.
+#[derive(Debug, PartialEq, Eq, Hash, Clone)]
+enum DistinctScalarValue {
+    Int8(i8),
+    Int16(i16),
+    Int32(i32),
+    Int64(i64),
+    UInt8(u8),
+    UInt16(u16),
+    UInt32(u32),
+    UInt64(u64),
+}
+
+/// For a given expression, maps its Arrow DataType into a LargeList of the
+/// same DataType.
+fn list_of(expr: Arc<dyn PhysicalExpr>, input_schema: &Schema) -> Result<DataType> {
+    let value_data_type = expr.data_type(input_schema)?;
+
+    match value_data_type {
+        DataType::Int8
+        | DataType::Int16
+        | DataType::Int32
+        | DataType::Int64
+        | DataType::UInt8
+        | DataType::UInt16
+        | DataType::UInt32
+        | DataType::UInt64 => Ok(DataType::LargeList(Box::new(value_data_type))),
+        _ => Err(ExecutionError::NotImplemented(
+            "Unsupported column data type for DISTINCT".to_string(),
+        )),
+    }
+}
+
+fn accumulate_scalar(
+    accum: &mut FnvHashSet<DistinctScalarValue>,
+    value: Option<ScalarValue>,
+) -> Result<()> {
+    let accum_value = match value {
+        Some(ScalarValue::Int8(v)) => Some(DistinctScalarValue::Int8(v)),

Review comment:
       ... this way, we do not need to go `ScalarValue -> DistinctScalarValue`...

##########
File path: rust/datafusion/src/physical_plan/distinct_expressions.rs
##########
@@ -0,0 +1,303 @@
+// 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.
+
+//! Implementations for DISTINCT expressions, e.g. `COUNT(DISTINCT c)`
+
+use std::cell::RefCell;
+use std::fmt::Debug;
+use std::hash::Hash;
+use std::rc::Rc;
+use std::sync::Arc;
+
+use arrow::array::ArrayRef;
+use arrow::array::{
+    Int16Array, Int32Array, Int64Array, Int8Array, PrimitiveArrayOps, UInt16Array,
+    UInt32Array, UInt64Array, UInt8Array,
+};
+use arrow::datatypes::{DataType, Schema};
+use arrow::record_batch::RecordBatch;
+
+use fnv::FnvHashSet;
+
+use crate::error::{ExecutionError, Result};
+use crate::logical_plan::ScalarValue;
+use crate::physical_plan::expressions::Column;
+use crate::physical_plan::hash_aggregate::AggregateMode;
+use crate::physical_plan::{Accumulator, AggregateExpr, PhysicalExpr};
+
+/// Enumeration of types that can be accumulated into a distinct set of values.
+#[derive(Debug, PartialEq, Eq, Hash, Clone)]
+enum DistinctScalarValue {
+    Int8(i8),
+    Int16(i16),
+    Int32(i32),
+    Int64(i64),
+    UInt8(u8),
+    UInt16(u16),
+    UInt32(u32),
+    UInt64(u64),
+}
+
+/// For a given expression, maps its Arrow DataType into a LargeList of the
+/// same DataType.
+fn list_of(expr: Arc<dyn PhysicalExpr>, input_schema: &Schema) -> Result<DataType> {
+    let value_data_type = expr.data_type(input_schema)?;
+
+    match value_data_type {
+        DataType::Int8
+        | DataType::Int16
+        | DataType::Int32
+        | DataType::Int64
+        | DataType::UInt8
+        | DataType::UInt16
+        | DataType::UInt32
+        | DataType::UInt64 => Ok(DataType::LargeList(Box::new(value_data_type))),
+        _ => Err(ExecutionError::NotImplemented(
+            "Unsupported column data type for DISTINCT".to_string(),
+        )),
+    }
+}
+
+fn accumulate_scalar(
+    accum: &mut FnvHashSet<DistinctScalarValue>,
+    value: Option<ScalarValue>,
+) -> Result<()> {
+    let accum_value = match value {
+        Some(ScalarValue::Int8(v)) => Some(DistinctScalarValue::Int8(v)),
+        Some(ScalarValue::Int16(v)) => Some(DistinctScalarValue::Int16(v)),
+        Some(ScalarValue::Int32(v)) => Some(DistinctScalarValue::Int32(v)),
+        Some(ScalarValue::Int64(v)) => Some(DistinctScalarValue::Int64(v)),
+        Some(ScalarValue::UInt8(v)) => Some(DistinctScalarValue::UInt8(v)),
+        Some(ScalarValue::UInt16(v)) => Some(DistinctScalarValue::UInt16(v)),
+        Some(ScalarValue::UInt32(v)) => Some(DistinctScalarValue::UInt32(v)),
+        Some(ScalarValue::UInt64(v)) => Some(DistinctScalarValue::UInt64(v)),
+        Some(ScalarValue::Null) => None,
+        _ => {
+            return Err(ExecutionError::NotImplemented(
+                "Unsupported scalar value for DISTINCT accumulator".to_string(),
+            ))
+        }
+    };
+
+    match accum_value {
+        Some(v) => {
+            accum.insert(v);
+        }
+        None => {}
+    }
+
+    Ok(())
+}
+
+macro_rules! accum_batch {
+    ($ARRAY_TY:ident, $DISTINCT_SCALAR_TY: path, $ARRAY: expr, $ACCUM: expr) => {{
+        let array = $ARRAY.as_any().downcast_ref::<$ARRAY_TY>().ok_or_else(|| {
+            ExecutionError::ExecutionError("Error downcasting array".to_string())
+        })?;
+
+        for i in 0..array.len() {
+            $ACCUM.insert($DISTINCT_SCALAR_TY(array.value(i)));
+        }
+
+        Ok(())
+    }};
+}
+
+#[derive(Debug)]
+struct DistinctValuesAccumulator {
+    values: FnvHashSet<DistinctScalarValue>,
+}
+
+impl Accumulator for DistinctValuesAccumulator {
+    fn accumulate_scalar(&mut self, value: Option<ScalarValue>) -> Result<()> {
+        accumulate_scalar(&mut self.values, value)
+    }
+
+    fn accumulate_batch(&mut self, _array: &ArrayRef) -> Result<()> {
+        Err(ExecutionError::NotImplemented(

Review comment:
       Wouldn't it be possible to implement this one here? Loop through the elements one by one.




----------------------------------------------------------------
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org