Removing datafusion-python as a rust dependency

Author: timsaucer

Cargo.lock

@@ -31,10 +31,6 @@ build = "build.rs"
 [dependencies]
 datafusion = { version = "42.0.0", features = ["pyarrow", "avro"] }
 datafusion-proto = "42.0.0"
-
-# temporarily point to revision until version 42 is released
-datafusion-python = { git = "https://github.com/apache/datafusion-python" }
-
 futures = "0.3"
 log = "0.4"
 prost = "0.13"

Cargo.toml

@@ -32,14 +32,14 @@ fn main() -> Result<(), String> {
 
     // We don't include the proto files in releases so that downstreams
     // do not need to have PROTOC included
-    if Path::new("src/proto/datafusion-ray.proto").exists() {
+    if Path::new("src/proto/datafusion_ray.proto").exists() {
         println!("cargo:rerun-if-changed=src/proto/datafusion.proto");
-        println!("cargo:rerun-if-changed=src/proto/datafusion-ray.proto");
+        println!("cargo:rerun-if-changed=src/proto/datafusion_ray.proto");
         tonic_build::configure()
             .extern_path(".datafusion", "::datafusion_proto::protobuf")
-            .compile(&["src/proto/datafusion-ray.proto"], &["src/proto"])
+            .compile(&["src/proto/datafusion_ray.proto"], &["src/proto"])
             .map_err(|e| format!("protobuf compilation failed: {e}"))?;
-        let generated_source_path = out.join("datafusion-ray.protobuf.rs");
+        let generated_source_path = out.join("datafusion_ray.protobuf.rs");
         let code = std::fs::read_to_string(generated_source_path).unwrap();
         let mut file = std::fs::OpenOptions::new()
             .write(true)

build.rs

@@ -25,8 +25,8 @@
     ExecutionGraph,
     QueryStage,
     execute_partition,
-    serialize_execution_plan,
-    deserialize_execution_plan,
+    # serialize_execution_plan,
+    # deserialize_execution_plan,
 )
 from .context import DatafusionRayContext
 

datafusion_ray/__init__.py

@@ -15,10 +15,12 @@
 # specific language governing permissions and limitations
 # under the License.
 
-import pytest
 from datafusion_ray import Context
+from datafusion import SessionContext
+
 
 def test():
-    ctx = Context(1)
-    ctx.register_csv('tips', 'examples/tips.csv', True)
+    df_ctx = SessionContext()
+    ctx = Context(df_ctx, False)
+    df_ctx.register_csv("tips", "examples/tips.csv", has_header=True)
     ctx.plan("SELECT * FROM tips")

datafusion_ray/tests/test_context.py

@@ -36,7 +36,6 @@ use datafusion_proto::bytes::{
 };
 use datafusion_proto::physical_plan::{AsExecutionPlan, DefaultPhysicalExtensionCodec};
 use datafusion_proto::protobuf;
-use datafusion_python::physical_plan::PyExecutionPlan;
 use futures::StreamExt;
 use prost::{DecodeError, Message};
 use pyo3::exceptions::PyRuntimeError;
@@ -54,6 +53,26 @@ pub struct PyContext {
     pub(crate) py_ctx: PyObject,
 }
 
+pub(crate) fn execution_plan_from_pyany(
+    py_plan: &Bound<PyAny>,
+) -> PyResult<Arc<dyn ExecutionPlan>> {
+    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 codec = DefaultPhysicalExtensionCodec {};
+    let runtime = RuntimeEnv::default();
+    let registry = SessionContext::new();
+    plan_node
+        .try_into_physical_plan(&registry, &runtime, &codec)
+        .map_err(|e| e.into())
+}
+
 #[pymethods]
 impl PyContext {
     #[new]
@@ -114,20 +133,9 @@ impl PyContext {
         // 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_proto = py_plan.call_method0(py, "to_proto")?;
-        let plan_bytes: &[u8] = py_proto.extract(py)?;
-        let plan_node = protobuf::PhysicalPlanNode::decode(plan_bytes).map_err(|e| {
-            PyRuntimeError::new_err(format!(
-                "Unable to decode physical plan protobuf message: {}",
-                e
-            ))
-        })?;
-
-        let codec = DefaultPhysicalExtensionCodec {};
-        let runtime = RuntimeEnv::default();
-        let registry = SessionContext::new();
-        let plan = plan_node.try_into_physical_plan(&registry, &runtime, &codec)?;
+        let py_plan = py_plan.bind(py);
 
+        let plan = execution_plan_from_pyany(py_plan)?;
         let graph = make_execution_graph(plan.clone())?;
 
         // debug logging
@@ -147,7 +155,7 @@ impl PyContext {
     /// 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: PyExecutionPlan,
+        plan: &Bound<'_, PyAny>,
         part: usize,
         inputs: PyObject,
         py: Python,
@@ -158,7 +166,7 @@ impl PyContext {
 
 #[pyfunction]
 pub fn execute_partition(
-    plan: PyExecutionPlan,
+    plan: &Bound<'_, PyAny>,
     part: usize,
     inputs: PyObject,
     py: Python,
@@ -171,20 +179,20 @@ pub fn execute_partition(
 }
 
 // TODO(@lsf) change this to use pickle
-#[pyfunction]
-pub fn serialize_execution_plan(plan: PyExecutionPlan) -> PyResult<Vec<u8>> {
-    let codec = ShuffleCodec {};
-    Ok(physical_plan_to_bytes_with_extension_codec(plan.plan, &codec)?.to_vec())
-}
+// #[pyfunction]
+// pub fn serialize_execution_plan(plan: Py<PyAny>) -> PyResult<Vec<u8>> {
+//     let codec = ShuffleCodec {};
+//     Ok(physical_plan_to_bytes_with_extension_codec(plan.plan, &codec)?.to_vec())
+// }
 
-#[pyfunction]
-pub fn deserialize_execution_plan(bytes: Vec<u8>) -> PyResult<PyExecutionPlan> {
-    let ctx = SessionContext::new();
-    let codec = ShuffleCodec {};
-    Ok(PyExecutionPlan::new(
-        physical_plan_from_bytes_with_extension_codec(&bytes, &ctx, &codec)?,
-    ))
-}
+// #[pyfunction]
+// pub fn deserialize_execution_plan(bytes: Vec<u8>) -> PyResult<PyExecutionPlan> {
+//     let ctx = SessionContext::new();
+//     let codec = ShuffleCodec {};
+//     Ok(PyExecutionPlan::new(
+//         physical_plan_from_bytes_with_extension_codec(&bytes, &ctx, &codec)?,
+//     ))
+// }
 
 /// Iterate down an ExecutionPlan and set the input objects for RayShuffleReaderExec.
 fn _set_inputs_for_ray_shuffle_reader(
@@ -235,7 +243,7 @@ fn _set_inputs_for_ray_shuffle_reader(
 /// 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: PyExecutionPlan,
+    py_plan: &Bound<'_, PyAny>,
     part: usize,
     inputs: PyObject,
 ) -> Result<Vec<RecordBatch>> {
@@ -248,19 +256,21 @@ fn _execute_partition(
         HashMap::new(),
         Arc::new(RuntimeEnv::default()),
     ));
+
+    let plan = execution_plan_from_pyany(py_plan)
+        .map_err(|e| DataFusionError::Execution(e.to_string()))?;
     Python::with_gil(|py| {
         let input_partitions = inputs
-            .bind(py)
-            .downcast::<PyList>()
+            .downcast_bound::<PyList>(py)
             .map_err(|e| DataFusionError::Execution(format!("{}", e)))?;
-        _set_inputs_for_ray_shuffle_reader(plan.plan.clone(), input_partitions)
+        _set_inputs_for_ray_shuffle_reader(plan.clone(), input_partitions)
     })?;
 
     // 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.plan.execute(part, ctx)?;
+        let mut stream = plan.execute(part, ctx)?;
         let mut results = vec![];
         while let Some(result) = stream.next().await {
             results.push(result?);

src/context.rs

@@ -20,7 +20,8 @@ extern crate core;
 use pyo3::prelude::*;
 
 mod proto;
-use crate::context::{deserialize_execution_plan, execute_partition, serialize_execution_plan};
+// use crate::context::{deserialize_execution_plan, execute_partition, serialize_execution_plan};
+use crate::context::execute_partition;
 pub use proto::generated::protobuf;
 
 pub mod context;
@@ -37,7 +38,7 @@ fn _datafusion_ray_internal(m: &Bound<'_, PyModule>) -> PyResult<()> {
     m.add_class::<planner::PyExecutionGraph>()?;
     m.add_class::<query_stage::PyQueryStage>()?;
     m.add_function(wrap_pyfunction!(execute_partition, m)?)?;
-    m.add_function(wrap_pyfunction!(serialize_execution_plan, m)?)?;
-    m.add_function(wrap_pyfunction!(deserialize_execution_plan, m)?)?;
+    // m.add_function(wrap_pyfunction!(serialize_execution_plan, m)?)?;
+    // m.add_function(wrap_pyfunction!(deserialize_execution_plan, m)?)?;
     Ok(())
 }

src/lib.rs

@@ -20,7 +20,6 @@ use datafusion::error::Result;
 use datafusion::physical_plan::{ExecutionPlan, Partitioning};
 use datafusion::prelude::SessionContext;
 use datafusion_proto::bytes::physical_plan_from_bytes_with_extension_codec;
-use datafusion_python::physical_plan::PyExecutionPlan;
 use pyo3::prelude::*;
 use std::sync::Arc;
 
@@ -51,9 +50,9 @@ impl PyQueryStage {
         self.stage.id
     }
 
-    pub fn get_execution_plan(&self) -> PyExecutionPlan {
-        PyExecutionPlan::new(self.stage.plan.clone())
-    }
+    // pub fn get_execution_plan(&self) -> PyExecutionPlan {
+    //     PyExecutionPlan::new(self.stage.plan.clone())
+    // }
 
     pub fn get_child_stage_ids(&self) -> Vec<usize> {
         self.stage.get_child_stage_ids()