Pulse Documentation

name: op-synth-monotonic-from
description: Strictly increasing or decreasing integer counter; deterministic per row, ignores RNG. Synthetic primary keys.
kind: operator
category: SYNTH
operator: monotonic_from
type: reference
applies_to: inspect, predict, manifest
examples_tags: [synth, cohort-analysis]

Synth distributions emit per-row values; no Response.Components.

Use when

Counts up (or down) by a fixed step from a start value, one step per row, to make unique IDs or row numbers.

Questions it answers:

• How do I give each synthetic row a unique customer ID?
• How do I number rows from 1000 upward?

Use something else:

• constant when every row should get the same value.
• uniform when the numbers should be random, not in order.
• regex when IDs should be text codes like AB-1234.

Params

• start — int, default 0. First emitted value.
• step — int, default 1, non-zero. Negative ⇒ decreasing.

The sampler holds cur = start - step and adds step before emitting, so row 0 emits start.

Inputs

Field type: — numeric u4/u8/u16/u32/u64, f32/f64. u64 for primary keys.

Output

Per-row float64 start + i * step for row i. RNG state untouched, so seeded streams stay independent of monotonic columns.

Gotchas

• step == 0 → SERVICE_VALIDATION at spec parse.
• Consumes NO RNG — adding / removing a monotonic_from field is the only spec edit preserving byte-equality of every other field’s stream.
• Rejected rows still advance the counter, so a post-constraint cohort has gaps. For dense IDs, sanitize constraints.
• Casting overflow is SILENT — start=250, step=10, row_count=10 overflows u8 after row 1. Declare wide enough, or use u64.
• Pairs with pulse synth from-schema for fixtures; Seed is irrelevant here.

See

• pulse_examples_search tags=[synth]
• Skills: synthetic-data, op-synth-constant, op-synth-uniform-date