datafusion_catalog/
streaming.rs

1// Licensed to the Apache Software Foundation (ASF) under one
2// or more contributor license agreements.  See the NOTICE file
3// distributed with this work for additional information
4// regarding copyright ownership.  The ASF licenses this file
5// to you under the Apache License, Version 2.0 (the
6// "License"); you may not use this file except in compliance
7// with the License.  You may obtain a copy of the License at
8//
9//   http://www.apache.org/licenses/LICENSE-2.0
10//
11// Unless required by applicable law or agreed to in writing,
12// software distributed under the License is distributed on an
13// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
14// KIND, either express or implied.  See the License for the
15// specific language governing permissions and limitations
16// under the License.
17
18//! A simplified [`TableProvider`] for streaming partitioned datasets
19
20use std::sync::Arc;
21
22use arrow::datatypes::SchemaRef;
23use async_trait::async_trait;
24use datafusion_common::{DFSchema, Result, plan_err};
25use datafusion_expr::{Expr, SortExpr, TableType};
26use datafusion_physical_expr::equivalence::project_ordering;
27use datafusion_physical_expr::{LexOrdering, create_physical_sort_exprs};
28use datafusion_physical_plan::ExecutionPlan;
29use datafusion_physical_plan::streaming::{PartitionStream, StreamingTableExec};
30use log::debug;
31
32use crate::{Session, TableProvider};
33
34/// A [`TableProvider`] that streams a set of [`PartitionStream`]
35#[derive(Debug)]
36pub struct StreamingTable {
37    schema: SchemaRef,
38    partitions: Vec<Arc<dyn PartitionStream>>,
39    infinite: bool,
40    sort_order: Vec<SortExpr>,
41}
42
43impl StreamingTable {
44    /// Try to create a new [`StreamingTable`] returning an error if the schema is incorrect
45    pub fn try_new(
46        schema: SchemaRef,
47        partitions: Vec<Arc<dyn PartitionStream>>,
48    ) -> Result<Self> {
49        for x in partitions.iter() {
50            let partition_schema = x.schema();
51            if !schema.contains(partition_schema) {
52                debug!(
53                    "target schema does not contain partition schema. \
54                        Target_schema: {schema:?}. Partition Schema: {partition_schema:?}"
55                );
56                return plan_err!("Mismatch between schema and batches");
57            }
58        }
59
60        Ok(Self {
61            schema,
62            partitions,
63            infinite: false,
64            sort_order: vec![],
65        })
66    }
67
68    /// Sets streaming table can be infinite.
69    pub fn with_infinite_table(mut self, infinite: bool) -> Self {
70        self.infinite = infinite;
71        self
72    }
73
74    /// Sets the existing ordering of streaming table.
75    pub fn with_sort_order(mut self, sort_order: Vec<SortExpr>) -> Self {
76        self.sort_order = sort_order;
77        self
78    }
79}
80
81#[async_trait]
82impl TableProvider for StreamingTable {
83    fn schema(&self) -> SchemaRef {
84        Arc::clone(&self.schema)
85    }
86
87    fn table_type(&self) -> TableType {
88        TableType::View
89    }
90
91    async fn scan(
92        &self,
93        state: &dyn Session,
94        projection: Option<&Vec<usize>>,
95        _filters: &[Expr],
96        limit: Option<usize>,
97    ) -> Result<Arc<dyn ExecutionPlan>> {
98        let physical_sort = if !self.sort_order.is_empty() {
99            let df_schema = DFSchema::try_from(Arc::clone(&self.schema))?;
100            let eqp = state.execution_props();
101
102            let original_sort_exprs =
103                create_physical_sort_exprs(&self.sort_order, &df_schema, eqp)?;
104
105            if let Some(p) = projection {
106                // When performing a projection, the output columns will not match
107                // the original physical sort expression indices. Also the sort columns
108                // may not be in the output projection. To correct for these issues
109                // we need to project the ordering based on the output schema.
110                let schema = Arc::new(self.schema.project(p)?);
111                LexOrdering::new(original_sort_exprs)
112                    .and_then(|lex_ordering| project_ordering(&lex_ordering, &schema))
113                    .map(|lex_ordering| lex_ordering.to_vec())
114                    .unwrap_or_default()
115            } else {
116                original_sort_exprs
117            }
118        } else {
119            vec![]
120        };
121
122        Ok(Arc::new(StreamingTableExec::try_new(
123            Arc::clone(&self.schema),
124            self.partitions.clone(),
125            projection,
126            LexOrdering::new(physical_sort),
127            self.infinite,
128            limit,
129        )?))
130    }
131}
datafusion_catalog/streaming.rs

datafusion_catalog/
streaming.rs
