DeltaCommand¶

DeltaCommand is a marker interface for delta commands.

Implementations¶

• AlterDeltaTableCommand
• ConvertToDeltaCommand
• DeleteCommand
• MergeIntoCommand
• OptimizeExecutor
• OptimizeTableCommandBase
• RestoreTableCommand
• StatisticsCollection
• UpdateCommand
• VacuumCommandImpl
• WriteIntoDelta

Converting Predicate Text to Catalyst Expression¶

parsePredicates(
  spark: SparkSession,
  predicate: String): Seq[Expression]

parsePredicates converts the given predicate text to an Expression (Spark SQL).

parsePredicates requests the given SparkSession (Spark SQL) for the session ParserInterface (Spark SQL) to parseExpression the given predicate text.

parsePredicates is used when:

• OptimizeTableCommand is executed (to convert the partitionPredicate)
• WriteIntoDelta command is requested to write (to convert the replaceWhere option with predicates)

Verifying Partition Predicates¶

verifyPartitionPredicates(
  spark: SparkSession,
  partitionColumns: Seq[String],
  predicates: Seq[Expression]): Unit

verifyPartitionPredicates asserts that the given predicates expressions are as follows:

• Contain no subqueries
• Reference partition columns only (partitionColumns)

verifyPartitionPredicates is used when:

• OptimizeTableCommand is executed (to verify the partitionPredicate if defined)
• WriteIntoDelta command is requested to write (to verify the replaceWhere option for SaveMode.Overwrite mode)
• StatisticsCollection utility is used to recompute statistics of a delta table

generateCandidateFileMap¶

generateCandidateFileMap(
  basePath: Path,
  candidateFiles: Seq[AddFile]): Map[String, AddFile]

generateCandidateFileMap...FIXME

generateCandidateFileMap is used when...FIXME

removeFilesFromPaths¶

removeFilesFromPaths(
  deltaLog: DeltaLog,
  nameToAddFileMap: Map[String, AddFile],
  filesToRewrite: Seq[String],
  operationTimestamp: Long): Seq[RemoveFile]

removeFilesFromPaths...FIXME

removeFilesFromPaths is used when:

• DeleteCommand and UpdateCommand commands are executed

Creating HadoopFsRelation (with TahoeBatchFileIndex)¶

buildBaseRelation(
  spark: SparkSession,
  txn: OptimisticTransaction,
  actionType: String,
  rootPath: Path,
  inputLeafFiles: Seq[String],
  nameToAddFileMap: Map[String, AddFile]): HadoopFsRelation

buildBaseRelation converts the given inputLeafFiles to AddFiles (with the given rootPath and nameToAddFileMap).

buildBaseRelation creates a TahoeBatchFileIndex for the AddFiles (with the input actionType and rootPath).

In the end, buildBaseRelation creates a HadoopFsRelation (Spark SQL) with the TahoeBatchFileIndex (and the other properties based on the metadata of the given OptimisticTransaction).

buildBaseRelation is used when:

• DeleteCommand and UpdateCommand commands are executed (with delete and update action types, respectively)

getTouchedFile¶

getTouchedFile(
  basePath: Path,
  filePath: String,
  nameToAddFileMap: Map[String, AddFile]): AddFile

getTouchedFile...FIXME

getTouchedFile is used when:

• DeltaCommand is requested to removeFilesFromPaths and create a HadoopFsRelation (for DeleteCommand and UpdateCommand commands)

• MergeIntoCommand is executed

isCatalogTable¶

isCatalogTable(
  analyzer: Analyzer,
  tableIdent: TableIdentifier): Boolean

isCatalogTable...FIXME

isCatalogTable is used when:

• ConvertToDeltaCommandBase is requested to isCatalogTable

isPathIdentifier¶

isPathIdentifier(
  tableIdent: TableIdentifier): Boolean

isPathIdentifier...FIXME

isPathIdentifier is used when...FIXME

commitLarge¶

commitLarge(
  spark: SparkSession,
  txn: OptimisticTransaction,
  actions: Iterator[Action],
  op: DeltaOperations.Operation,
  context: Map[String, String],
  metrics: Map[String, String]): Long

commitLarge...FIXME

commitLarge is used when:

• ConvertToDeltaCommand command is executed (and requested to performConvert)
• RestoreTableCommand command is executed

updateAndCheckpoint¶

updateAndCheckpoint(
  spark: SparkSession,
  deltaLog: DeltaLog,
  commitSize: Int,
  attemptVersion: Long): Unit

updateAndCheckpoint requests the given DeltaLog to update.

updateAndCheckpoint prints out the following INFO message to the logs:

Committed delta #[attemptVersion] to [logPath]. Wrote [commitSize] actions.

In the end, updateAndCheckpoint requests the given DeltaLog to checkpoint the current snapshot.

IllegalStateException¶

updateAndCheckpoint throws an IllegalStateException if the version after update does not match the assumed attemptVersion:

The committed version is [attemptVersion] but the current version is [currentSnapshot].

Logging¶

DeltaCommand is an abstract class and logging is configured using the logger of the implementations.