multi-project

The multi-project operator expands each input row into multiple output rows using a set of templated projections. It is useful when you need to fan out one record into several derived records while keeping full control over the output schema.

Syntax

| multi-project ( <column-name> : <column-type> [, ...] ) [ <column-expr>, ... ]

Parameters

• <column-name>: The name of an output column.
• <column-type>: The data type of the output column (e.g., string, int64, datetime).
• <column-expr>: An expression that produces the value for the corresponding column. Expressions may reference columns from the input stream or constants.

Return Value

For each input row, multi-project outputs one row per templated set of expressions. The resulting table has only the columns defined in the multi-project column list.

Notes

• The number of expressions in each templated row must match the number of columns declared.
• Each templated row is evaluated independently per input row; if you define N templates and the input has M rows, the output will have M × N rows.
• Column types are validated against the expressions at parse time.

Examples

Example 1: Expand each row into multiple stages

datatable (User:string, EventCount:int64)
[
    "alice", 3,
    "bob",   2
]
| multi-project (Stage:string, CountDelta:int64) [
    "pre",  EventCount - 1,
    "post", EventCount + 1
]

Stage	CountDelta
pre	2
post	4
pre	1
post	3

Example 2: Duplicate rows with transformed values

datatable (Timestamp:datetime, Message:string)
[
    datetime(2024-01-01), "Started"
]
| multi-project (Event:string, Occurred:datetime) [
    strcat("raw_", Message), Timestamp,
    strcat("normalized_", tolower(Message)), Timestamp + time(1h)
]

This produces two rows per input record: one with the original message and one with a normalized variant an hour later.