LogicalPlan

type LogicalPlan =
  | ReadPlan
  | ReadTablePlan
  | LocalRelationPlan
  | SqlPlan
  | FilterPlan
  | ProjectPlan
  | AggregatePlan
  | LimitPlan
  | SortPlan
  | JoinPlan
  | DropPlan
  | WithColumnsPlan
  | DeduplicatePlan
  | OffsetPlan
  | CatalogPlan
  | SetOperationPlan
  | SamplePlan
  | NAFillPlan
  | NADropPlan
  | ToDFPlan
  | DescribePlan
  | RangePlan
  | WithColumnsRenamedPlan
  | SubqueryAliasPlan
  | HintPlan
  | TailPlan
  | RepartitionPlan
  | RepartitionByExpressionPlan
  | UnpivotPlan
  | SummaryPlan
  | NAReplacePlan
  | StatCorrPlan
  | StatCovPlan
  | StatCrosstabPlan
  | StatFreqItemsPlan
  | StatApproxQuantilePlan;

Defined in: plan/logical-plan.ts:94

The logical plan tree, mirroring Spark Catalyst’s internal representation. Every DataFrame transformation appends a new node to the tree; the full tree serializes to protobuf for Spark Connect when an action runs.

The plan that spark-connect-js builds is the unresolved form: column references are strings, nothing is bound to a catalog yet. The Spark Connect server runs the remaining Catalyst phases (analysis, optimization, physical planning, codegen) before executing.

A malformed unresolved plan produces cryptic AnalysisExceptions on the server, so the shape here matters. Each variant maps one-to-one to a Spark Connect Relation protobuf message.

LogicalPlan

See