`random`

Generic Description

Is the random-value alias for ad hoc nondeterministic sampling. It is documented separately because numeric transforms change ranking, thresholds, and buckets in ways planner tools must not guess.

Simple example:

RETURN random() AS value

Consumer-Level Explanation

Use it when the alias reads better in exploratory queries, with the same reproducibility caveats as rand. Prefer this function when the transformation is a deliberate part of the model or report, and name the derived value in a WITH stage when it will be reused.

More Detailed Explanation

random is scalar math, so it is evaluated per row after graph patterns, procedure output, or map projections have produced values. Keep it close to the query when thresholds, ranking, and derived fields need to be auditable and reproducible.

Advanced Example

This example applies random to a named score component after matching graph data, making the derived metric reviewable before ordering.

MATCH (d:Document)
WITH d, random() AS sample_score
RETURN d.title AS document, sample_score
ORDER BY sample_score
LIMIT 2

Real Use Cases

• distance-from-target scoring
• numeric bucketing and normalization before ORDER BY or aggregation
• making scoring formulas visible to query review and assistant tooling

Real Limitations And Tradeoffs

• invalid numeric inputs produce null-like results rather than a repaired value
• math functions do not fix poor feature scaling; choose transformations intentionally
• nondeterministic functions such as random sampling are not appropriate for persisted decisions