Dataframes

Adapters return QueryResult, containing one or more typed dataframes. Fields are columnar: every field has exactly frame.length values.

import type { QueryResult } from "@openplait/core";

const result: QueryResult = {
  frames: [{
    name: "requests-by-service",
    length: 2,
    fields: [
      { name: "service", type: "string", values: ["checkout", "payments"] },
      { name: "requests", type: "number", values: [1842, 907] },
    ],
  }],
  metadata: {
    datasource: "primary",
    rowsReturned: 2,
    executionTimeMs: 18,
    cache: "miss",
    warnings: [],
  },
};

Field types include strings, numbers, booleans, times, durations, traces, logs, JSON, and other portable visualization-friendly values. Optional units describe how a value should be formatted without coupling it to a chart library.

Use validateQueryResult(result) at integration boundaries. It checks field shape, lengths, names, and metadata invariants before product code consumes a third-party adapter response.

Transformations

The runtime can filter, select, rename, calculate, sort, limit, group, join, merge, pivot, unpivot, convert units, fill missing values, apply windows, expand labels, and extract JSON properties. Placement can be source, runtime, client, or capability-aware auto.