Time-series in PLOMID — overview, v0.1.0 docs

Note

No dedicated time-series engine in v0.1.0: no hypertables, continuous aggregates, retention, or downsampling. Time-series = temporal columns + SQL + ts indexes + columnar DeltaBitpack/Rle + zone-map/BRIN pruning. Anything beyond that is roadmap.

Query patternsWindows, rollups, FILTER aggregates, keyset feeds Temporal referenceLiterals, arithmetic, EXTRACT, precision

Model#

sqlsource

CREATE TABLE events (
  id BIGINT PRIMARY KEY,
  ts TIMESTAMPTZ NOT NULL,
  kind TEXT NOT NULL,
  payload JSONB
);
CREATE INDEX events_ts_idx ON events (ts);

One row per event, ts indexed, hot JSON dimensions in payload. Columnar flush encodes timestamp runs (DeltaBitpack) and sorted keys (Rle); scans prune by ZoneMap → BRIN → XOR → Roaring → exact.

diagram

flowchart LR
    I["INSERT event"] --> IDX["B-tree ts index"]
    I --> COL["columnar flush\nDeltaBitpack timestamps"]
    Q["time-window query"] --> PR["ZoneMap → BRIN → XOR → Roaring → exact"]
    PR --> R["matching rows"]

Diagram source · mermaidcopy included

mermaidsource

flowchart LR
    I["INSERT event"] --> IDX["B-tree ts index"]
    I --> COL["columnar flush\nDeltaBitpack timestamps"]
    Q["time-window query"] --> PR["ZoneMap → BRIN → XOR → Roaring → exact"]
    PR --> R["matching rows"]

What works / what doesn't#

Need	v0.1.0
Insert events, query windows, order, aggregate	Supported
`date_trunc` rollups, `FILTER` splits	Supported
Retention/downsampling/continuous agg	Not available — roll up on read or delete old partitions manually
Gap-filling / interpolation	Not available — outer-join a `generate_series` calendar (see queries)

Query time guide · Indexes · Columnar

Time-series in PLOMID — overview

Model#

What works / what doesn't#

Related#

Continue