Skip to main content

Parquet: Values From Columns

Extract several columns from one Parquet row in a single pass. Cheaper and clearer than wiring N separate per-column nodes when you need many fields off the same row.

Each configured column adds a dedicated output connector — one per requested column — typed according to its Value type.

Parameters​

Skip on error

Shown as "Ignore errors" in the editor. When enabled, a record that fails (missing column or type mismatch on a mandatory entry) is marked as skippable downstream — the record is soft-skipped instead of failing the run.

Columns

One entry per column to extract (the list itself is optional). Each entry has:

  • Label — display label shown next to the connector in the editor.
  • Column — source column name in the Parquet row.
  • Value type — expected PValue tag of the cell (String, Number, Bool, …).
  • Mandatory — rows added in the editor start with this off. When enabled, a missing column or type mismatch aborts the whole node. When disabled, only that connector is left empty and the row continues without emitting a value for this column.

Each entry's internal uuid is its output connector key: adding / removing entries changes the node's output connectors, and removing a row drops downstream links pointing at its key.

Input​

Parquet rowrequired

One row from a Parquet stream (typically produced by a Parquet reader upstream).

Output​

One output connector per configured column (keyed by the entry's internal uuid), typed according to that column's Value type. The connector is labeled "{label} ({column}) ({type})", falling back to a translated "{column} ({type})" template when the Label is empty.