context.rs

// 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 crate::ext::Extensions;
use crate::planner::{make_execution_graph, PyExecutionGraph};
use datafusion::arrow::pyarrow::ToPyArrow;
use datafusion::arrow::record_batch::RecordBatch;
use datafusion::error::{DataFusionError, Result};
use datafusion::execution::context::TaskContext;
use datafusion::execution::runtime_env::RuntimeEnv;
use datafusion::physical_plan::{displayable, ExecutionPlan};
use datafusion::prelude::*;
use datafusion_proto::physical_plan::AsExecutionPlan;
use datafusion_proto::protobuf;
use datafusion_python::physical_plan::PyExecutionPlan;
use futures::StreamExt;
use prost::Message;
use pyo3::exceptions::PyRuntimeError;
use pyo3::prelude::*;
use pyo3::types::{PyBytes, PyTuple};
use std::collections::HashMap;
use std::sync::Arc;
use tokio::runtime::Runtime;
use tokio::task::JoinHandle;

type PyResultSet = Vec<PyObject>;

#[pyclass(name = "Context", module = "datafusion_ray", subclass)]
pub struct PyContext {
    pub(crate) py_ctx: PyObject,
}

pub(crate) fn execution_plan_from_pyany(
    py_plan: &Bound<PyAny>,
    py: Python,
) -> PyResult<Arc<dyn ExecutionPlan>> {
    if let Ok(py_plan) = py_plan.to_object(py).downcast_bound::<PyExecutionPlan>(py) {
        // For session contexts created with datafusion_ray.extended_session_context(), the inner
        // execution plan can be used as such (and the enabled extensions are all available).
        Ok(py_plan.borrow().plan.clone())
    } else {
        // The session context originates from outside our library, so we'll grab the protobuf plan
        // by calling the python method with no extension codecs.
        let py_proto = py_plan.call_method0("to_proto")?;
        let plan_bytes: &[u8] = py_proto.extract()?;
        let plan_node = protobuf::PhysicalPlanNode::try_decode(plan_bytes).map_err(|e| {
            PyRuntimeError::new_err(format!(
                "Unable to decode physical plan protobuf message: {}",
                e
            ))
        })?;

        let runtime = RuntimeEnv::default();
        let registry = SessionContext::new();
        plan_node
            .try_into_physical_plan(&registry, &runtime, Extensions::codec())
            .map_err(|e| e.into())
    }
}

#[pymethods]
impl PyContext {
    #[new]
    pub fn new(session_ctx: PyObject) -> Result<Self> {
        Ok(Self {
            py_ctx: session_ctx,
        })
    }

    /// Execute SQL directly against the DataFusion context. Useful for statements
    /// such as "create view" or "drop view"
    pub fn sql(&self, query: &str, py: Python) -> PyResult<()> {
        println!("Executing {}", query);
        // let _df = wait_for_future(py, self.ctx.sql(sql))?;
        let _df = self.run_sql(query, py);
        Ok(())
    }

    fn run_sql(&self, query: &str, py: Python) -> PyResult<Py<PyAny>> {
        let args = PyTuple::new_bound(py, [query]);
        self.py_ctx.call_method1(py, "sql", args)
    }

    /// Plan a distributed SELECT query for executing against the Ray workers
    pub fn plan(&self, plan: &Bound<PyAny>, py: Python) -> PyResult<PyExecutionGraph> {
        // println!("Planning {}", sql);
        // let df = wait_for_future(py, self.ctx.sql(sql))?;
        // let py_df = self.run_sql(sql, py)?;
        // let py_plan = py_df.call_method0(py, "execution_plan")?;
        // let py_plan = py_plan.bind(py);

        let plan = execution_plan_from_pyany(plan, py)?;
        let graph = make_execution_graph(plan.clone())?;

        // debug logging
        let mut stages = graph.query_stages.values().collect::<Vec<_>>();
        stages.sort_by_key(|s| s.id);
        for stage in stages {
            println!(
                "Query stage #{}:\n{}",
                stage.id,
                displayable(stage.plan.as_ref()).indent(false)
            );
        }

        Ok(PyExecutionGraph::new(graph))
    }

    /// Execute a partition of a query plan. This will typically be executing a shuffle write and write the results to disk
    pub fn execute_partition(
        &self,
        plan: &Bound<'_, PyBytes>,
        part: usize,
        py: Python,
    ) -> PyResult<PyResultSet> {
        execute_partition(plan, part, py)
    }
}

#[pyfunction]
pub fn execute_partition(
    plan_bytes: &Bound<'_, PyBytes>,
    part: usize,
    py: Python,
) -> PyResult<PyResultSet> {
    let plan = deserialize_execution_plan(plan_bytes)?;
    _execute_partition(plan, part)
        .unwrap()
        .into_iter()
        .map(|batch| batch.to_pyarrow(py))
        .collect()
}

pub fn serialize_execution_plan(
    plan: Arc<dyn ExecutionPlan>,
    py: Python,
) -> PyResult<Bound<'_, PyBytes>> {
    let proto = datafusion_proto::protobuf::PhysicalPlanNode::try_from_physical_plan(
        plan.clone(),
        Extensions::codec(),
    )?;

    let bytes = proto.encode_to_vec();
    Ok(PyBytes::new_bound(py, &bytes))
}

pub fn deserialize_execution_plan(proto_msg: &Bound<PyBytes>) -> PyResult<Arc<dyn ExecutionPlan>> {
    let bytes: &[u8] = proto_msg.extract()?;
    let proto_plan =
        datafusion_proto::protobuf::PhysicalPlanNode::try_decode(bytes).map_err(|e| {
            PyRuntimeError::new_err(format!(
                "Unable to decode logical node from serialized bytes: {}",
                e
            ))
        })?;

    let ctx = SessionContext::new();
    let plan = proto_plan
        .try_into_physical_plan(&ctx, &ctx.runtime_env(), Extensions::codec())
        .map_err(DataFusionError::from)?;

    Ok(plan)
}

/// Execute a partition of a query plan. This will typically be executing a shuffle write and
/// write the results to disk, except for the final query stage, which will return the data.
/// inputs is a list of tuples of (stage_id, partition_id, bytes) for each input partition.
fn _execute_partition(plan: Arc<dyn ExecutionPlan>, part: usize) -> Result<Vec<RecordBatch>> {
    let ctx = Arc::new(TaskContext::new(
        Some("task_id".to_string()),
        "session_id".to_string(),
        SessionConfig::default(),
        HashMap::new(),
        HashMap::new(),
        HashMap::new(),
        Arc::new(RuntimeEnv::default()),
    ));

    // create a Tokio runtime to run the async code
    let rt = Runtime::new().unwrap();

    let fut: JoinHandle<Result<Vec<RecordBatch>>> = rt.spawn(async move {
        let mut stream = plan.execute(part, ctx)?;
        let mut results = vec![];
        while let Some(result) = stream.next().await {
            results.push(result?);
        }
        Ok(results)
    });

    // block and wait on future
    let results = rt.block_on(fut).unwrap()?;
    Ok(results)
}