One in, one out
A job is atomic. Several sources or several destinations is the role of a workflow, not of pipeline.
A job reads from exactly one source and writes to exactly one destination. pipeline.yaml names them.
pipeline: from: src_input to: dest_output
Both names must resolve to something declared, and in the right direction.
Written by hand: this block has no Pydantic model in the engine, it is parsed directly by the runner.
| key | type | required | default | constraint |
|---|---|---|---|---|
| from | string | yes | — | an id declared in sources |
| to | string | yes | — | an id declared in destinations |
A job is atomic. Several sources or several destinations is the role of a workflow, not of pipeline.
from and to hold declared identifiers. The file path lives in extract.table and load.table.