[GitHub] [arrow-datafusion] Jimexist commented on a change in pull request #1561: add correlation function

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

Jimexist commented on a change in pull request #1561:
URL: https://github.com/apache/arrow-datafusion/pull/1561#discussion_r785383000



##########
File path: datafusion/src/physical_plan/expressions/correlation.rs
##########
@@ -0,0 +1,546 @@
+// 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.
+
+//! Defines physical expressions that can evaluated at runtime during query execution
+
+use std::any::Any;
+use std::sync::Arc;
+
+use crate::error::{DataFusionError, Result};
+use crate::physical_plan::{
+    expressions::{covariance::CovarianceAccumulator, stddev::StddevAccumulator},
+    Accumulator, AggregateExpr, PhysicalExpr,
+};
+use crate::scalar::ScalarValue;
+use arrow::{array::ArrayRef, datatypes::DataType, datatypes::Field};
+
+use super::{format_state_name, StatsType};
+
+/// CORR aggregate expression
+#[derive(Debug)]
+pub struct Correlation {
+    name: String,
+    expr1: Arc<dyn PhysicalExpr>,
+    expr2: Arc<dyn PhysicalExpr>,
+}
+
+/// function return type of correlation
+pub(crate) fn correlation_return_type(arg_type: &DataType) -> Result<DataType> {
+    match arg_type {
+        DataType::Int8
+        | DataType::Int16
+        | DataType::Int32
+        | DataType::Int64
+        | DataType::UInt8
+        | DataType::UInt16
+        | DataType::UInt32
+        | DataType::UInt64
+        | DataType::Float32
+        | DataType::Float64 => Ok(DataType::Float64),
+        other => Err(DataFusionError::Plan(format!(
+            "CORR does not support {:?}",
+            other
+        ))),
+    }
+}
+
+pub(crate) fn is_correlation_support_arg_type(arg_type: &DataType) -> bool {
+    matches!(
+        arg_type,
+        DataType::UInt8
+            | DataType::UInt16
+            | DataType::UInt32
+            | DataType::UInt64
+            | DataType::Int8
+            | DataType::Int16
+            | DataType::Int32
+            | DataType::Int64
+            | DataType::Float32
+            | DataType::Float64
+    )
+}
+
+impl Correlation {
+    /// Create a new COVAR_POP aggregate function
+    pub fn new(
+        expr1: Arc<dyn PhysicalExpr>,
+        expr2: Arc<dyn PhysicalExpr>,
+        name: impl Into<String>,
+        data_type: DataType,
+    ) -> Self {
+        // the result of correlation just support FLOAT64 data type.
+        assert!(matches!(data_type, DataType::Float64));
+        Self {
+            name: name.into(),
+            expr1,
+            expr2,
+        }
+    }
+}
+
+impl AggregateExpr for Correlation {
+    /// Return a reference to Any that can be used for downcasting
+    fn as_any(&self) -> &dyn Any {
+        self
+    }
+
+    fn field(&self) -> Result<Field> {
+        Ok(Field::new(&self.name, DataType::Float64, true))
+    }
+
+    fn create_accumulator(&self) -> Result<Box<dyn Accumulator>> {
+        Ok(Box::new(CorrelationAccumulator::try_new()?))
+    }
+
+    fn state_fields(&self) -> Result<Vec<Field>> {
+        Ok(vec![
+            Field::new(
+                &format_state_name(&self.name, "count"),
+                DataType::UInt64,
+                true,
+            ),
+            Field::new(
+                &format_state_name(&self.name, "mean1"),
+                DataType::Float64,
+                true,
+            ),
+            Field::new(
+                &format_state_name(&self.name, "m2_1"),
+                DataType::Float64,
+                true,
+            ),
+            Field::new(
+                &format_state_name(&self.name, "mean2"),
+                DataType::Float64,
+                true,
+            ),
+            Field::new(
+                &format_state_name(&self.name, "m2_2"),
+                DataType::Float64,
+                true,
+            ),
+            Field::new(
+                &format_state_name(&self.name, "algo_const"),
+                DataType::Float64,
+                true,
+            ),
+        ])
+    }
+
+    fn expressions(&self) -> Vec<Arc<dyn PhysicalExpr>> {
+        vec![self.expr1.clone(), self.expr2.clone()]
+    }
+
+    fn name(&self) -> &str {
+        &self.name
+    }
+}
+
+/// An accumulator to compute correlation
+#[derive(Debug)]
+pub struct CorrelationAccumulator {
+    covar: CovarianceAccumulator,
+    stddev1: StddevAccumulator,
+    stddev2: StddevAccumulator,
+}
+
+impl CorrelationAccumulator {
+    /// Creates a new `CorrelationAccumulator`
+    pub fn try_new() -> Result<Self> {
+        Ok(Self {
+            covar: CovarianceAccumulator::try_new(StatsType::Population)?,
+            stddev1: StddevAccumulator::try_new(StatsType::Population)?,
+            stddev2: StddevAccumulator::try_new(StatsType::Population)?,
+        })
+    }
+}
+
+impl Accumulator for CorrelationAccumulator {
+    fn state(&self) -> Result<Vec<ScalarValue>> {
+        Ok(vec![
+            ScalarValue::from(self.covar.get_count()),
+            ScalarValue::from(self.covar.get_mean1()),
+            ScalarValue::from(self.stddev1.get_m2()),
+            ScalarValue::from(self.covar.get_mean2()),
+            ScalarValue::from(self.stddev2.get_m2()),
+            ScalarValue::from(self.covar.get_algo_const()),
+        ])
+    }
+
+    fn update_batch(&mut self, values: &[ArrayRef]) -> Result<()> {
+        self.covar.update_batch(values)?;
+        self.stddev1.update_batch(&[values[0].clone()])?;
+        self.stddev2.update_batch(&[values[1].clone()])?;
+        Ok(())
+    }
+
+    fn merge_batch(&mut self, states: &[ArrayRef]) -> Result<()> {
+        let states_c = [
+            states[0].clone(),
+            states[1].clone(),
+            states[3].clone(),
+            states[5].clone(),
+        ];
+        let states_s1 = [states[0].clone(), states[1].clone(), states[2].clone()];
+        let states_s2 = [states[0].clone(), states[3].clone(), states[4].clone()];
+
+        self.covar.merge_batch(&states_c)?;
+        self.stddev1.merge_batch(&states_s1)?;
+        self.stddev2.merge_batch(&states_s2)?;
+        Ok(())
+    }
+
+    fn update(&mut self, _values: &[ScalarValue]) -> Result<()> {
+        unimplemented!("update_batch is implemented instead");
+    }
+
+    fn merge(&mut self, _states: &[ScalarValue]) -> Result<()> {
+        unimplemented!("merge_batch is implemented instead");
+    }
+
+    fn evaluate(&self) -> Result<ScalarValue> {
+        let covar = self.covar.evaluate()?;
+        let stddev1 = self.stddev1.evaluate()?;
+        let stddev2 = self.stddev2.evaluate()?;
+
+        if let ScalarValue::Float64(Some(c)) = covar {
+            if let ScalarValue::Float64(Some(s1)) = stddev1 {
+                if let ScalarValue::Float64(Some(s2)) = stddev2 {
+                    return Ok(ScalarValue::Float64(Some(c / s1 / s2)));

Review comment:
       can s1 or s2 be 0?




-- 
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: github-unsubscribe@arrow.apache.org

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