Skip to main content

SAMPLE BY

Downsampling with explicit fill semantics — the query-time twin of the storage fill policies.

SELECT ts, <agg>(<channel>), ... FROM <table>
  WHERE entity = '...' AND ts BETWEEN <a> AND <b>
  SAMPLE BY <duration> [FILL(PREV|NULL|LINEAR)];

Rules:

  • requires a single entity and a bounded ts range
  • every selected channel column must be aggregated (bare ts and entity are fine)
  • aggregates and SAMPLE BY come together or not at all

Aggregates​

FunctionInputResult
count(c)anyint
avg(c), sum(c), min(c), max(c)f64f64
first(c), last(c)anychannel type

FILL modes​

Buckets with no samples are, by mode:

  • (no FILL) — absent from the result
  • FILL(NULL) — present, aggregate cells null
  • FILL(PREV) — carry the previous bucket's value (hold, at bucket granularity)
  • FILL(LINEAR) — linear interpolation between the nearest non-empty neighbors (numeric only; edges stay null)
SELECT ts, avg(speed) FROM telemetry
  WHERE entity = 'veh-1' AND ts BETWEEN 1700000000000 AND 1700086400000
  SAMPLE BY 5 m FILL(LINEAR);

Row ts is the bucket start (buckets align to floor(from / interval)).