Pulse Documentation

name: op-win-lag
description: Per-row value of Field from N rows earlier in the ordered partition.
kind: operator
category: WIN
operator: WIN_LAG
type: reference
applies_to: process, compose, predict
examples_tags: [time-series, window-operator, buffered-pipeline]

Window operators emit row-level values; they do not produce Response.Components.

Use when

Adds to every row the value of a field from a set number of rows earlier in the ordered series, such as last month’s sales.

Questions it answers:

• What were each store’s sales in the month before, on the same row as this month’s?
• What score did each respondent give in the previous wave?

Use something else:

• WIN_DELTA when you want the difference from the earlier row, not its value.

Params

partition_by (carve), order_by (≥1, required, numeric / date), frame (forbidden).

Inputs

Output

One float64 per row written to Label (default WIN_LAG_<field>). When i - offset < 0 within the partition emits default if set, else null.

Gotchas

• Not weightable: a request weight is PROCESSING_CONFIG (windows carry no slot weight); Options.DefaultWeight is skipped.
• order_by required — predict rejects empty slate (PULSE_WINDOW_INVALID).
• frame forbidden — set one and predict rejects.
• Partitioning by the raw date collapses each row to its own partition. Pick a coarser partition (region, account) and order by date.
• Forces buffered execution (Streamable=false).

See

• pulse_examples_search tags=[time-series]
• Skills: window-design, op-win-lead, op-win-pct-change