Trait datafusion_catalog::TableProvider

source ·
pub trait TableProvider: Sync + Send {
    // Required methods
    fn as_any(&self) -> &dyn Any;
    fn schema(&self) -> SchemaRef;
    fn table_type(&self) -> TableType;
    fn scan<'life0, 'life1, 'life2, 'life3, 'async_trait>(
        &'life0 self,
        state: &'life1 dyn Session,
        projection: Option<&'life2 Vec<usize>>,
        filters: &'life3 [Expr],
        limit: Option<usize>,
    ) -> Pin<Box<dyn Future<Output = Result<Arc<dyn ExecutionPlan>>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait,
             'life1: 'async_trait,
             'life2: 'async_trait,
             'life3: 'async_trait;

    // Provided methods
    fn constraints(&self) -> Option<&Constraints> { ... }
    fn get_table_definition(&self) -> Option<&str> { ... }
    fn get_logical_plan(&self) -> Option<&LogicalPlan> { ... }
    fn get_column_default(&self, _column: &str) -> Option<&Expr> { ... }
    fn supports_filters_pushdown(
        &self,
        filters: &[&Expr],
    ) -> Result<Vec<TableProviderFilterPushDown>> { ... }
    fn statistics(&self) -> Option<Statistics> { ... }
    fn insert_into<'life0, 'life1, 'async_trait>(
        &'life0 self,
        _state: &'life1 dyn Session,
        _input: Arc<dyn ExecutionPlan>,
        _overwrite: bool,
    ) -> Pin<Box<dyn Future<Output = Result<Arc<dyn ExecutionPlan>>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait,
             'life1: 'async_trait { ... }
}
Expand description

Source table

Required Methods§

source

fn as_any(&self) -> &dyn Any

Returns the table provider as Any so that it can be downcast to a specific implementation.

source

fn schema(&self) -> SchemaRef

Get a reference to the schema for this table

source

fn table_type(&self) -> TableType

Get the type of this table for metadata/catalog purposes.

source

fn scan<'life0, 'life1, 'life2, 'life3, 'async_trait>( &'life0 self, state: &'life1 dyn Session, projection: Option<&'life2 Vec<usize>>, filters: &'life3 [Expr], limit: Option<usize>, ) -> Pin<Box<dyn Future<Output = Result<Arc<dyn ExecutionPlan>>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait, 'life3: 'async_trait,

Create an ExecutionPlan for scanning the table with optionally specified projection, filter and limit, described below.

The ExecutionPlan is responsible scanning the datasource’s partitions in a streaming, parallelized fashion.

§Projection

If specified, only a subset of columns should be returned, in the order specified. The projection is a set of indexes of the fields in Self::schema.

DataFusion provides the projection to scan only the columns actually used in the query to improve performance, an optimization called “Projection Pushdown”. Some datasources, such as Parquet, can use this information to go significantly faster when only a subset of columns is required.

§Filters

A list of boolean filter Exprs to evaluate during the scan, in the manner specified by Self::supports_filters_pushdown. Only rows for which all of the Exprs evaluate to true must be returned (aka the expressions are ANDed together).

To enable filter pushdown you must override Self::supports_filters_pushdown as the default implementation does not and filters will be empty.

DataFusion pushes filtering into the scans whenever possible (“Filter Pushdown”), and depending on the format and the implementation of the format, evaluating the predicate during the scan can increase performance significantly.

§Note: Some columns may appear only in Filters

In certain cases, a query may only use a certain column in a Filter that has been completely pushed down to the scan. In this case, the projection will not contain all the columns found in the filter expressions.

For example, given the query SELECT t.a FROM t WHERE t.b > 5,

┌────────────────────┐
│  Projection(t.a)   │
└────────────────────┘
           ▲
           │
           │
┌────────────────────┐     Filter     ┌────────────────────┐   Projection    ┌────────────────────┐
│  Filter(t.b > 5)   │────Pushdown──▶ │  Projection(t.a)   │ ───Pushdown───▶ │  Projection(t.a)   │
└────────────────────┘                └────────────────────┘                 └────────────────────┘
           ▲                                     ▲                                      ▲
           │                                     │                                      │
           │                                     │                           ┌────────────────────┐
┌────────────────────┐                ┌────────────────────┐                 │        Scan        │
│        Scan        │                │        Scan        │                 │  filter=(t.b > 5)  │
└────────────────────┘                │  filter=(t.b > 5)  │                 │  projection=(t.a)  │
                                      └────────────────────┘                 └────────────────────┘

Initial Plan                  If `TableProviderFilterPushDown`           Projection pushdown notes that
                              returns true, filter pushdown              the scan only needs t.a
                              pushes the filter into the scan
                                                                         BUT internally evaluating the
                                                                         predicate still requires t.b
§Limit

If limit is specified, must only produce at least this many rows, (though it may return more). Like Projection Pushdown and Filter Pushdown, DataFusion pushes LIMITs as far down in the plan as possible, called “Limit Pushdown” as some sources can use this information to improve their performance. Note that if there are any Inexact filters pushed down, the LIMIT cannot be pushed down. This is because inexact filters do not guarantee that every filtered row is removed, so applying the limit could lead to too few rows being available to return as a final result.

Provided Methods§

source

fn constraints(&self) -> Option<&Constraints>

Get a reference to the constraints of the table. Returns:

  • None for tables that do not support constraints.
  • Some(&Constraints) for tables supporting constraints. Therefore, a Some(&Constraints::empty()) return value indicates that this table supports constraints, but there are no constraints.
source

fn get_table_definition(&self) -> Option<&str>

Get the create statement used to create this table, if available.

source

fn get_logical_plan(&self) -> Option<&LogicalPlan>

Get the LogicalPlan of this table, if available

source

fn get_column_default(&self, _column: &str) -> Option<&Expr>

Get the default value for a column, if available.

source

fn supports_filters_pushdown( &self, filters: &[&Expr], ) -> Result<Vec<TableProviderFilterPushDown>>

Specify if DataFusion should provide filter expressions to the TableProvider to apply during the scan.

Some TableProviders can evaluate filters more efficiently than the Filter operator in DataFusion, for example by using an index.

§Parameters and Return Value

The return Vec must have one element for each element of the filters argument. The value of each element indicates if the TableProvider can apply the corresponding filter during the scan. The position in the return value corresponds to the expression in the filters parameter.

If the length of the resulting Vec does not match the filters input an error will be thrown.

Each element in the resulting Vec is one of the following:

  • Exact or Inexact: The TableProvider can apply the filter during scan
  • Unsupported: The TableProvider cannot apply the filter during scan

By default, this function returns Unsupported for all filters, meaning no filters will be provided to Self::scan.

§Example
// Define a struct that implements the TableProvider trait
struct TestDataSource {}

#[async_trait]
impl TableProvider for TestDataSource {
        todo!()
    // Override the supports_filters_pushdown to evaluate which expressions
    // to accept as pushdown predicates.
    fn supports_filters_pushdown(&self, filters: &[&Expr]) -> Result<Vec<TableProviderFilterPushDown>> {
        // Process each filter
        let support: Vec<_> = filters.iter().map(|expr| {
          match expr {
            // This example only supports a between expr with a single column named "c1".
            Expr::Between(between_expr) => {
                between_expr.expr
                .try_into_col()
                .map(|column| {
                    if column.name == "c1" {
                        TableProviderFilterPushDown::Exact
                    } else {
                        TableProviderFilterPushDown::Unsupported
                    }
                })
                // If there is no column in the expr set the filter to unsupported.
                .unwrap_or(TableProviderFilterPushDown::Unsupported)
            }
            _ => {
                // For all other cases return Unsupported.
                TableProviderFilterPushDown::Unsupported
            }
        }
    }).collect();
    Ok(support)
    }
}
source

fn statistics(&self) -> Option<Statistics>

Get statistics for this table, if available

source

fn insert_into<'life0, 'life1, 'async_trait>( &'life0 self, _state: &'life1 dyn Session, _input: Arc<dyn ExecutionPlan>, _overwrite: bool, ) -> Pin<Box<dyn Future<Output = Result<Arc<dyn ExecutionPlan>>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Return an ExecutionPlan to insert data into this table, if supported.

The returned plan should return a single row in a UInt64 column called “count” such as the following

+-------+,
| count |,
+-------+,
| 6     |,
+-------+,
§See Also

See DataSinkExec for the common pattern of inserting a streams of RecordBatches as files to an ObjectStore.

Implementors§