Catalog

Data engineering & pipelines

Tasks around ingestion, transformation, and orchestration. The Oracle produces the exact expected dataset, and validation compares schema and row-level output deterministically.

Example onlyTerminal Benchsqlsparketl
Task promptFix the incremental load so late-arriving events are deduplicated correctly.
Reference approach (illustrative)An illustrative approach: define a stable event identity and an explicit late-arrival policy, then test deduplication with late-event fixtures. Exact keys depend on the schema.