date_range validator

Faithful summary of lib/validators.ts. Validators are pure functions and run identically in the worker, the Workbench, and the MCP server.

function validDateRange(value, params) {
  const t = Date.parse(value)
  if (Number.isNaN(t)) return fail('date_invalid', 'not a parseable date')
  const min = params.min === 'today' ? Date.now()
            : params.min ? Date.parse(params.min) : -Infinity
  const max = params.max === 'today' ? Date.now()
            : params.max ? Date.parse(params.max) : Infinity
  if (t < min) return fail('date_too_early', `before ${params.min}`)
  if (t > max) return fail('date_too_late',  `after ${params.max}`)
  return OK
}

Validators attach to a field on a custom template. The cascade runs them after mapping but before commit; failures surface in the per-row validation report.

fields:
  - column: date_of_birth
    type: date
    format: auto
    validators:
      - type: date_range
        min: '1900-01-01'
        max: today

• Parsing is engine-dependent. `Date.parse("13/06/2024")` returns NaN in V8. Use ISO `YYYY-MM-DD` or pre-format in your transform.
• `today` resolves at validation time, not request time — there is no caching across rows in a batch.
• Time zone: bare dates parse as UTC midnight. A `min: today` boundary in non-UTC zones can reject rows that "look" current locally.

• Patient demographicspatient_demographics_v1

  Patient identity. Emits FHIR Patient.

  Fields: date_of_birth

• KYC profileskyc_profiles_v1

  Know-Your-Customer identity profiles with risk rating. Contains PII.

  Fields: date_of_birth

Validators add no AI cost — they are pure functions. Every map bills a small flat per-map fee from your prepaid token wallet ($10 minimum, shared across our suite); only columns that reach the metered LLM layer, plus any-to-any convert and reshape, bill real AI tokens. See pricing →

Back to the documentation index or browse the template catalogue.