Package org.apache.iceberg.spark

Class RollbackStagedTable

java.lang.Object
org.apache.iceberg.spark.RollbackStagedTable
All Implemented Interfaces:
org.apache.spark.sql.connector.catalog.StagedTable, org.apache.spark.sql.connector.catalog.SupportsDelete, org.apache.spark.sql.connector.catalog.SupportsDeleteV2, org.apache.spark.sql.connector.catalog.SupportsRead, org.apache.spark.sql.connector.catalog.SupportsWrite, org.apache.spark.sql.connector.catalog.Table, org.apache.spark.sql.connector.catalog.TruncatableTable
public class RollbackStagedTable extends Object implements org.apache.spark.sql.connector.catalog.StagedTable, org.apache.spark.sql.connector.catalog.SupportsRead, org.apache.spark.sql.connector.catalog.SupportsWrite, org.apache.spark.sql.connector.catalog.SupportsDelete
An implementation of StagedTable that mimics the behavior of Spark's non-atomic CTAS and RTAS.

A Spark catalog can implement StagingTableCatalog to support atomic operations by producing StagedTable. But if a catalog implements StagingTableCatalog, Spark expects the catalog to be able to produce a StagedTable for any table loaded by the catalog. This assumption doesn't always work, as in the case of SparkSessionCatalog, which supports atomic operations can produce a StagedTable for Iceberg tables, but wraps the session catalog and cannot necessarily produce a working StagedTable implementation for tables that it loads.

The work-around is this class, which implements the StagedTable interface but does not have atomic behavior. Instead, the StagedTable interface is used to implement the behavior of the non-atomic SQL plans that will create a table, write, and will drop the table to roll back.

This StagedTable implements SupportsRead, SupportsWrite, and SupportsDelete by passing the calls to the real table. Implementing those interfaces is safe because Spark will not use them unless the table supports them and returns the corresponding capabilities from capabilities().

  • Constructor Summary

    Constructors
    Constructor
    Description
    RollbackStagedTable(org.apache.spark.sql.connector.catalog.TableCatalog catalog, org.apache.spark.sql.connector.catalog.Identifier ident, org.apache.spark.sql.connector.catalog.Table table)
     

  • Method Summary

    Modifier and Type
    Method
    Description
    void
    abortStagedChanges()
     
    Set<org.apache.spark.sql.connector.catalog.TableCapability>
    capabilities()
     
    void
    commitStagedChanges()
     
    void
    deleteWhere(org.apache.spark.sql.sources.Filter[] filters)
     
    String
    name()
     
    org.apache.spark.sql.connector.read.ScanBuilder
    newScanBuilder(org.apache.spark.sql.util.CaseInsensitiveStringMap options)
     
    org.apache.spark.sql.connector.write.WriteBuilder
    newWriteBuilder(org.apache.spark.sql.connector.write.LogicalWriteInfo info)
     
    org.apache.spark.sql.connector.expressions.Transform[]
    partitioning()
     
    Map<String,String>
    properties()
     
    org.apache.spark.sql.types.StructType
    schema()
     

    Methods inherited from class java.lang.Object

    clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

    Methods inherited from interface org.apache.spark.sql.connector.catalog.SupportsDelete

    canDeleteWhere, canDeleteWhere, deleteWhere, truncateTable

    Methods inherited from interface org.apache.spark.sql.connector.catalog.Table

    columns

  • Constructor Details

    • RollbackStagedTable

      public RollbackStagedTable(org.apache.spark.sql.connector.catalog.TableCatalog catalog, org.apache.spark.sql.connector.catalog.Identifier ident, org.apache.spark.sql.connector.catalog.Table table)

  • Method Details

    • commitStagedChanges

      public void commitStagedChanges()
      Specified by:
      commitStagedChanges in interface org.apache.spark.sql.connector.catalog.StagedTable

    • abortStagedChanges

      public void abortStagedChanges()
      Specified by:
      abortStagedChanges in interface org.apache.spark.sql.connector.catalog.StagedTable

    • name

      public String name()
      Specified by:
      name in interface org.apache.spark.sql.connector.catalog.Table

    • schema

      public org.apache.spark.sql.types.StructType schema()
      Specified by:
      schema in interface org.apache.spark.sql.connector.catalog.Table

    • partitioning

      public org.apache.spark.sql.connector.expressions.Transform[] partitioning()
      Specified by:
      partitioning in interface org.apache.spark.sql.connector.catalog.Table

    • properties

      public Map<String,String> properties()
      Specified by:
      properties in interface org.apache.spark.sql.connector.catalog.Table

    • capabilities

      public Set<org.apache.spark.sql.connector.catalog.TableCapability> capabilities()
      Specified by:
      capabilities in interface org.apache.spark.sql.connector.catalog.Table

    • deleteWhere

      public void deleteWhere(org.apache.spark.sql.sources.Filter[] filters)
      Specified by:
      deleteWhere in interface org.apache.spark.sql.connector.catalog.SupportsDelete

    • newScanBuilder

      public org.apache.spark.sql.connector.read.ScanBuilder newScanBuilder(org.apache.spark.sql.util.CaseInsensitiveStringMap options)
      Specified by:
      newScanBuilder in interface org.apache.spark.sql.connector.catalog.SupportsRead

    • newWriteBuilder

      public org.apache.spark.sql.connector.write.WriteBuilder newWriteBuilder(org.apache.spark.sql.connector.write.LogicalWriteInfo info)
      Specified by:
      newWriteBuilder in interface org.apache.spark.sql.connector.catalog.SupportsWrite