chore: Prepare for DataFusion 48.0.0 (#1710)

Author: andygrove

native/Cargo.lock

@@ -34,11 +34,11 @@ edition = "2021"
 rust-version = "1.85"
 
 [workspace.dependencies]
-arrow = { version = "55.0.0", features = ["prettyprint", "ffi", "chrono-tz"] }
+arrow = { version = "55.1.0", features = ["prettyprint", "ffi", "chrono-tz"] }
 async-trait = { version = "0.1" }
 bytes = { version = "1.10.0" }
-parquet = { version = "55.0.0", default-features = false, features = ["experimental"] }
-datafusion = { version = "47.0.0", default-features = false, features = ["unicode_expressions", "crypto_expressions", "nested_expressions", "parquet"] }
+parquet = { version = "55.1.0", default-features = false, features = ["experimental"] }
+datafusion = { git = "https://github.com/apache/datafusion", rev = "efd9587", default-features = false, features = ["unicode_expressions", "crypto_expressions", "nested_expressions", "parquet"] }
 datafusion-comet-spark-expr = { path = "spark-expr" }
 datafusion-comet-proto = { path = "proto" }
 chrono = { version = "0.4", default-features = false, features = ["clock"] }

native/Cargo.toml

@@ -81,7 +81,7 @@ jni = { version = "0.21", features = ["invocation"] }
 lazy_static = "1.4"
 assertables = "9"
 hex = "0.4.3"
-datafusion-functions-nested = { version = "47.0.0" }
+datafusion-functions-nested = { git = "https://github.com/apache/datafusion", rev = "efd9587" }
 
 [features]
 default = []

native/core/Cargo.toml

@@ -147,8 +147,8 @@ impl ExecutionPlan for CopyExec {
         )))
     }
 
-    fn statistics(&self) -> DataFusionResult<Statistics> {
-        self.input.statistics()
+    fn partition_statistics(&self, partition: Option<usize>) -> DataFusionResult<Statistics> {
+        self.input.partition_statistics(partition)
     }
 
     fn properties(&self) -> &PlanProperties {

native/core/src/execution/operators/copy.rs

@@ -166,7 +166,7 @@ impl FilterExec {
         predicate: &Arc<dyn PhysicalExpr>,
         default_selectivity: u8,
     ) -> Result<Statistics> {
-        let input_stats = input.statistics()?;
+        let input_stats = input.partition_statistics(None)?;
         let schema = input.schema();
         if !check_support(predicate, &schema) {
             let selectivity = default_selectivity as f64 / 100.0;

native/core/src/execution/operators/filter.rs

@@ -78,7 +78,7 @@ use datafusion::common::{
 use datafusion::datasource::listing::PartitionedFile;
 use datafusion::logical_expr::type_coercion::other::get_coerce_type_for_case_expression;
 use datafusion::logical_expr::{
-    AggregateUDF, ReturnTypeArgs, WindowFrame, WindowFrameBound, WindowFrameUnits,
+    AggregateUDF, ReturnFieldArgs, WindowFrame, WindowFrameBound, WindowFrameUnits,
     WindowFunctionDefinition,
 };
 use datafusion::physical_expr::expressions::{Literal, StatsType};
@@ -886,7 +886,7 @@ impl PhysicalPlanner {
                     func_name,
                     fun_expr,
                     vec![left, right],
-                    data_type,
+                    Field::new(func_name, data_type, true),
                 )))
             }
             _ => Ok(Arc::new(BinaryExpr::new(left, op, right))),
@@ -2191,6 +2191,12 @@ impl PhysicalPlanner {
                         .coerce_types(&input_expr_types)
                         .unwrap_or_else(|_| input_expr_types.clone());
 
+                    let arg_fields = coerced_types
+                        .iter()
+                        .enumerate()
+                        .map(|(i, dt)| Field::new(format!("arg{i}"), dt.clone(), true))
+                        .collect::<Vec<_>>();
+
                     // TODO this should try and find scalar
                     let arguments = args
                         .iter()
@@ -2202,18 +2208,16 @@ impl PhysicalPlanner {
                         })
                         .collect::<Vec<_>>();
 
-                    let nullables = arguments.iter().map(|_| true).collect::<Vec<_>>();
-
-                    let args = ReturnTypeArgs {
-                        arg_types: &coerced_types,
+                    let args = ReturnFieldArgs {
+                        arg_fields: &arg_fields,
                         scalar_arguments: &arguments,
-                        nullables: &nullables,
                     };
 
                     let data_type = func
                         .inner()
-                        .return_type_from_args(args)?
-                        .return_type()
+                        .return_field_from_args(args)?
+                        .clone()
+                        .data_type()
                         .clone();
 
                     (data_type, coerced_types)
@@ -2246,7 +2250,7 @@ impl PhysicalPlanner {
             fun_name,
             fun_expr,
             args.to_vec(),
-            data_type,
+            Field::new(fun_name, data_type, true),
         ));
 
         Ok(scalar_expr)

native/core/src/execution/planner.rs

@@ -141,7 +141,7 @@ impl ExecutionPlan for ShuffleWriterExec {
     }
 
     fn statistics(&self) -> Result<Statistics> {
-        self.input.statistics()
+        self.input.partition_statistics(None)
     }
 
     fn properties(&self) -> &PlanProperties {

native/core/src/execution/shuffle/shuffle_writer.rs

@@ -78,8 +78,8 @@ pub(crate) fn init_datasource_exec(
             ))
         });
 
-        if let (Some(filter), Some(data_schema)) = (cnf_data_filters, &data_schema) {
-            parquet_source = parquet_source.with_predicate(Arc::clone(data_schema), filter);
+        if let Some(filter) = cnf_data_filters {
+            parquet_source = parquet_source.with_predicate(filter);
         }
     }
 

native/core/src/parquet/parquet_exec.rs

@@ -20,6 +20,7 @@
 use crate::parquet::parquet_support::{spark_parquet_convert, SparkParquetOptions};
 use arrow::array::{new_null_array, RecordBatch, RecordBatchOptions};
 use arrow::datatypes::{Schema, SchemaRef};
+use datafusion::common::ColumnStatistics;
 use datafusion::datasource::schema_adapter::{SchemaAdapter, SchemaAdapterFactory, SchemaMapper};
 use datafusion::physical_plan::ColumnarValue;
 use std::sync::Arc;
@@ -226,6 +227,13 @@ impl SchemaMapper for SchemaMapping {
         let record_batch = RecordBatch::try_new_with_options(schema, cols, &options)?;
         Ok(record_batch)
     }
+
+    fn map_column_statistics(
+        &self,
+        _file_col_statistics: &[ColumnStatistics],
+    ) -> datafusion::common::Result<Vec<ColumnStatistics>> {
+        Ok(vec![])
+    }
 }
 
 #[cfg(test)]

native/core/src/parquet/schema_adapter.rs

@@ -38,7 +38,7 @@ use arrow::{
     record_batch::RecordBatch,
     util::display::FormatOptions,
 };
-use chrono::{NaiveDate, NaiveDateTime, TimeZone, Timelike};
+use chrono::{DateTime, NaiveDate, TimeZone, Timelike};
 use datafusion::common::{
     cast::as_generic_string_array, internal_err, Result as DataFusionResult, ScalarValue,
 };
@@ -2108,7 +2108,7 @@ fn date_parser(date_str: &str, eval_mode: EvalMode) -> SparkResult<Option<i32>>
     ) {
         Some(date) => {
             let duration_since_epoch = date
-                .signed_duration_since(NaiveDateTime::UNIX_EPOCH.date())
+                .signed_duration_since(DateTime::UNIX_EPOCH.naive_utc().date())
                 .num_days();
             Ok(Some(duration_since_epoch.to_i32().unwrap()))
         }