Language reference

Rand (random number)

Syntax

Rand()

π id, Rand() → r (Users)

Description

Rand returns a fresh random number between 0 (inclusive) and 1 (exclusive) each time it is called. Use it to shuffle, assign random buckets, or sample rows — Rand() < 0.1 keeps roughly 10% of rows.

Technical Description

Rand() → NUMBER in [0, 1). Zero-argument. It reads the system RNG, so it is neither PURE nor DETERMINISTIC — the optimizer never folds or dedupes it, and it produces a new value per evaluation. No seed (not reproducible). Underpins the SAMPLE p operator (which desugars to σ Rand() < p).

Examples

Randomly keep about 10% of rows:

σ Rand() < 0.1 (Events)

SELECT Rand() < 0.1 (Events)

Assign a random sort key (shuffle):

τ r ASC (π id, Rand() → r (Deck))

SORT r ASC (PROJECT id, Rand() -> r (Deck))

Random bucket 0–9:

π id, Int(Rand() * 10) → bucket (Users)

PROJECT id, Int(Rand() * 10) -> bucket (Users)

Limitations

Non-deterministic and unseeded — results differ every run and are not reproducible. Because it is impure it is never constant-folded.

Alternatives

SAMPLE p for probabilistic sampling and SAMPLE n ROWS for an exact-count sample express the common cases more directly.

Notes

Use Rand() with τ for a random shuffle, or with σ for ad-hoc sampling.