A government open-data team uses Cloud Data Fusion to onboard 40 source systems into BigQuery, each with its own schema, light cleansing rules, and sensitive fields that must be masked. Citizen analysts maintain the pipelines through a visual interface, while a separate platform team manages shared connections, credentials, and pipeline templates that all analyst pipelines reuse. Which Cloud Data Fusion capability lets the platform team package the shared connections and cleansing logic as reusable building blocks that analysts can drop into pipelines without writing code?
- AWrangler directives saved as ad hoc transformation scripts inside each individual pipeline by every analyst as they build it.
- BDataform repositories referenced from inside Cloud Data Fusion pipelines so each analyst pipeline calls a shared SQLX script for cleansing.
- CCloud Functions wrapped behind HTTP triggers that each pipeline must call for masking, registered as external services in the Data Fusion environment.
- DPlugins published from the Hub, combined with system-managed connections and reusable pipeline templates that analysts import into their own pipelines. Correct
Why A is wrong: Wrangler directives are powerful for interactive cleansing, but saving them inside each pipeline means logic is duplicated and cannot be governed centrally by the platform team; this misses the reuse requirement.
Why B is wrong: Dataform is a separate service that orchestrates SQL inside BigQuery; Cloud Data Fusion pipelines do not natively call SQLX scripts, so this is a misleading mix of services.
Why C is wrong: Calling Cloud Functions adds latency, cost, and operational overhead, and is not the idiomatic Data Fusion mechanism for reusable transforms; the visual-first model points to plugins and templates instead.
Why D is correct: Cloud Data Fusion plugins extend the visual palette with custom sources, transforms, and sinks; combined with named system connections and saved pipeline templates, they give analysts pre-blessed building blocks they drag into pipelines without writing code, which matches the team's requirements.