union
union [@dataset]+Combines the main pipeline input with every referenced additional dataset into one dataset whose rows are the union of all inputs.
Each argument must be a dataset reference (for example @other) naming a dataset other than the main input. Every input, including the main input, must share the same dataset kind: all Table, all Event, or all Interval. The main input cannot be a Resource.
Column names present in any input appear in the output. For a column missing from a given input, rows originating from that input carry null in that column. Types for the same column name must unify; when a wider numeric type is chosen, values may be cast with possible precision loss.
Valid-from and valid-to fields from Event and Interval inputs are renamed to match the main input’s time column names. Non-time columns on a side input that collide with the main valid-from or valid-to name are dropped from that side with a column-name collision warning.
Primary key, candidate keys, and column constness metadata are cleared on the output. The grouping key is kept only when every input’s grouping-key column set matches (order may differ). Time alignment is preserved only when all inputs share a compatible alignment; otherwise alignment is dropped.
Foreign keys from side inputs are merged in, with source fields rewritten through column renames.
Categories
Accelerable
union is always accelerable if the input is accelerable. A dataset that only uses accelerable verbs can be accelerated, making
queries on the dataset respond faster.
Examples
Stacks the main station rows with months so shared columns merge and missing fields become null on each side.
Swaps which dataset is the main input so the merged schema still lines up, with null filling columns that exist only on one side.