feat(events): schedule, recurrence and the calendar (Phase 4)
All checks were successful
PR Checks / bot-tests (pull_request) Successful in 37s
PR Checks / client-build (pull_request) Successful in 43s
PR Checks / server-tests (pull_request) Successful in 13m26s

The four closed recurrence shapes computed in the definition's own IANA zone,
a fourteen-day materialisation horizon with projections beyond it, series as a
managed thing, and the admin calendar that replaces the plugin this feature
exists to replace. An event now happens on its own.

No schema change: Phase 1 built every column this needed.

- events/recurrence.js is the ONE place an occurrence is computed, so the
  runner's expansion and the calendar's forecast cannot disagree. No date
  library added — Node ships the tzdata one would vendor, behind Intl.
- The runner's materialise leg is now two halves: expand, then sweep. The
  window starts at `now - grace`, so an occurrence nobody could have seen is
  never invented retroactively; the horizon is what makes the missed sweep
  mean anything for a recurrence.
- Publishing is the schedule switch and archiving turns it off, and publishing
  re-pins every occurrence that has not started.
- A projection is never drawn over an instant a run occupies, so a cancelled
  occurrence does not reappear as a forecast.

54 new tests, incl. the DST fixture set the plan asked for and three new
statements proved against a real MariaDB. Suite 1768/1711/56 skipped/1 fail
(pre-existing CRLF). Walked end to end on the local review stack.

Docs: RunicGateway/docs#PENDING

Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
2026-09-02 16:10:16 -05:00
parent a481248bc0
commit 6e73660b52
30 changed files with 3722 additions and 77 deletions

View File

@@ -0,0 +1,170 @@
// ── The calendar ───────────────────────────────────────────────────────────
//
// EVENTS.md §I: "month and list view, filtered by category, scope and series",
// and Phase 4's stated deliverable — *the thing this feature exists to replace*
// is a WordPress calendar plugin with no series field and no recurrence.
//
// **A calendar entry is one of two things, and the difference is not cosmetic.**
//
// - A **run**: a real `event_runs` row. It has an id, a status, a health, a
// pinned version and a console. Somebody can cancel it. It exists because the
// runner materialised it inside its fourteen-day horizon, or because an admin
// started it by hand.
// - A **projection**: arithmetic. There is no row, nothing to cancel, and
// nothing has been committed to. It exists so that a monthly event is visible
// three weeks out instead of the calendar simply ending at the horizon (org
// lead, 2026-09-02).
//
// The API says which each is and the UI renders them differently, because an
// operator acting on a projection as though it were a booking is the failure
// this distinction exists to prevent. A projection is a forecast of what the
// runner *will* materialise, computed by the same `occurrencesBetween` the
// runner itself calls — one arithmetic, so the forecast cannot disagree with
// what later appears.
//
// **A projection is never emitted for an instant a run already occupies**, which
// is what keeps the fortnight inside the horizon from showing everything twice.
// That rule also does the right thing for a CANCELLED occurrence: the row is
// still there, so nothing re-projects it, and an event an operator called off
// does not reappear on the calendar as though it were still coming.
const runsDb = require('./eventRuns.db')
const definitionsDb = require('./eventDefinitions.db')
const recurrence = require('../../events/recurrence')
// A calendar request is operator-supplied, and a year-wide window across forty
// weekly definitions is how a month view becomes an outage. Ninety-two days is
// a three-month view — more than the month grid and the list either need.
const MAX_WINDOW_DAYS = 92
const MAX_ENTRIES = 1000
const runEntry = (run) => ({
kind: 'run',
runId: run.id,
definitionId: run.definition_id,
title: run.definition_title,
slug: run.definition_slug,
seriesId: run.series_id || null,
seriesName: run.series_name || null,
seriesSlug: run.series_slug || null,
scheduledFor: run.scheduled_for,
timezone: run.timezone,
scope: run.scope,
status: run.status,
health: run.health,
version: run.version_number,
rehearsal: Boolean(run.rehearsal),
waitingSteps: Number(run.waiting_steps || 0),
})
const projectedEntry = (definition, occurrence) => ({
kind: 'projected',
runId: null,
definitionId: definition.id,
title: definition.title,
slug: definition.slug,
seriesId: definition.series_id || null,
seriesName: definition.series_name || null,
seriesSlug: definition.series_slug || null,
scheduledFor: occurrence.at,
timezone: definition.timezone,
scope: '',
status: null,
health: null,
// Why this instant is not the wall clock the schedule names. Carried on the
// projection as well as on the materialised run, so the calendar can explain
// a DST-shifted time before it happens rather than after.
adjusted: occurrence.adjusted,
shiftMinutes: occurrence.shiftMinutes,
})
/**
* The calendar for a window.
*
* `{ ok, window, horizon, entries }` — entries ascending by instant, runs and
* projections interleaved. `horizon` is the instant past which nothing is
* materialised yet, so the UI can draw the line rather than infer it.
*
* **The instants are UTC and the placement is the client's.** A month grid has
* one date axis and the viewer's own zone is what "this month" means to the
* person reading it; each entry carries its own `timezone` so the time beside it
* reads `20:00 Europe/Berlin` and nobody misreads a shard's local schedule as
* their own. That is the split §E's "the timezone belongs to the event" implies:
* the event owns the time, the reader owns the calendar.
*/
async function calendar({
from,
to,
status = null,
scope = null,
seriesId = null,
horizonDays = 14,
now = new Date(),
} = {}) {
const start = from instanceof Date ? from : new Date(from)
const end = to instanceof Date ? to : new Date(to)
if (Number.isNaN(start.getTime()) || Number.isNaN(end.getTime())) {
return { ok: false, status: 400, errors: ['from and to must be dates'] }
}
if (end <= start) {
return { ok: false, status: 400, errors: ['to must be after from'] }
}
if (end - start > MAX_WINDOW_DAYS * recurrence.DAY_MS) {
return { ok: false, status: 400, errors: [`the window may span at most ${MAX_WINDOW_DAYS} days`] }
}
const runs = await runsDb.listInWindow({ from: start, to: end, status, scope, seriesId })
const entries = runs.map(runEntry)
// Every instant a run already occupies, keyed by definition. Projections are
// per definition at the empty scope, so the definition and the instant are the
// whole key -- the same triple the unique index uses, with the scope fixed.
const taken = new Set(
runs
.filter((r) => !r.scope)
.map((r) => `${r.definition_id}@${new Date(r.scheduled_for).getTime()}`),
)
// A status filter is a filter on RUNS. A projection has no status, so asking
// for "everything that failed" must not answer with a forecast — it would be a
// forecast that failed, which is not a thing.
// The scope filter behaves the same way, and for the same reason: automatic
// expansion is at the empty scope (org lead, 2026-09-02), so a request narrowed
// to a named scope has no forecast to give.
if (!status && !scope) {
const definitions = await definitionsDb.findSchedulable()
for (const definition of definitions) {
if (seriesId && Number(definition.series_id) !== Number(seriesId)) continue
const schedule = definition.version_spec?.schedule
if (!schedule || schedule.kind === 'manual') continue
let occurrences = []
try {
occurrences = recurrence.occurrencesBetween(
schedule,
definition.timezone || 'UTC',
start,
end,
)
} catch {
continue
}
for (const occurrence of occurrences) {
if (taken.has(`${definition.id}@${occurrence.at.getTime()}`)) continue
entries.push(projectedEntry(definition, occurrence))
}
}
}
entries.sort((a, b) => new Date(a.scheduledFor) - new Date(b.scheduledFor))
return {
ok: true,
status: 200,
window: { from: start, to: end },
horizon: new Date(now.getTime() + horizonDays * recurrence.DAY_MS),
entries: entries.slice(0, MAX_ENTRIES),
truncated: entries.length > MAX_ENTRIES,
}
}
module.exports = { calendar, MAX_WINDOW_DAYS, MAX_ENTRIES }

View File

@@ -114,6 +114,39 @@ const markReady = (id, versionId, userId) =>
[versionId, userId, id],
)
/**
* Every definition the runner should expand a recurrence for (Phase 4).
*
* `ready` is the whole gate, and it is deliberately the only one: EVENTS.md §E
* defines `ready` as "a version has been published and the schedule is live", so
* publishing IS the switch and archiving is how an operator turns a recurrence
* off. A separate schedule-enabled flag would be a second answer to a question
* `state` already answers, and the two would eventually disagree.
*
* The VERSION's spec is joined rather than the definition's working copy: the
* draft is what an author is midway through editing, and a half-typed `weekly`
* must never materialise anything. The pinned spec comes back with it, so the
* whole expansion is one round trip.
*
* The series columns are here for the CALENDAR rather than the runner, which
* ignores them: a projected occurrence has to be filterable and labellable by
* its arc exactly as a materialised run is, and a second query to learn the name
* of a row this one already reached would be two round trips for a join.
*/
const findSchedulable = async () => {
const rows = await query(
`SELECT d.id, d.title, d.slug, d.timezone, d.grace_seconds, d.concurrency_key,
d.current_version_id, d.series_id, v.spec AS version_spec,
s.name AS series_name, s.slug AS series_slug
FROM event_definitions d
JOIN event_versions v ON v.id = d.current_version_id
LEFT JOIN event_series s ON s.id = d.series_id
WHERE d.state = 'ready'
ORDER BY d.id`,
)
return rows.map((row) => ({ ...row, version_spec: parseJson(row.version_spec, null) }))
}
/**
* Archive. Never a hard delete while runs reference it (§ API surface) — and the
* schema would refuse one anyway, because `event_runs.version_id` RESTRICTs.
@@ -127,6 +160,7 @@ module.exports = {
getById,
getBySlug,
slugTaken,
findSchedulable,
insert,
update,
markReady,

View File

@@ -23,6 +23,7 @@
const db = require('./eventDefinitions.db')
const versionsDb = require('./eventVersions.db')
const runsDb = require('./eventRuns.db')
const logDb = require('./eventRunLog.db')
const seriesDb = require('./eventSeries.db')
const spec = require('../../events/spec')
const { slugify, uniqueSlug } = require('../teams/teamSlug')
@@ -229,7 +230,31 @@ async function publish(id, userId) {
const version = await versionsDb.nextVersion(id)
const versionId = await versionsDb.insert(id, version, checked.spec, userId)
await db.markReady(id, versionId, userId)
return { ok: true, versionId, version, definition: await db.getById(id) }
// Occurrences already materialised ahead of their instant move to the new
// version; ones that have begun do not (org lead, 2026-09-02). Logged per run
// rather than only counted, because "which version did this run actually use"
// is the first question an audit asks and the pin is no longer immutable while
// a run is still `scheduled`.
const pending = await runsDb.listScheduledFor(id)
const stale = pending.filter((r) => Number(r.version_id) !== Number(versionId))
const repinned = stale.length ? await runsDb.repinScheduled(id, versionId) : 0
for (const run of stale) {
await logDb.write({
runId: run.id,
kind: 'run.status',
detail: {
to: 'scheduled',
repinned: true,
fromVersionId: run.version_id,
toVersionId: versionId,
toVersion: version,
by: userId,
},
})
}
return { ok: true, versionId, version, repinned, definition: await db.getById(id) }
}
/**

View File

@@ -97,6 +97,92 @@ const materialise = async (run) => {
return Number(result?.affectedRows || 0) === 1 ? result.insertId : null
}
/**
* Every run whose instant falls inside a window — the calendar's real half.
*
* Ascending, unlike the admin run list: a calendar is read forwards. The join
* reaches the series so a month can be filtered to one arc without a second
* round trip, and `d.timezone` is NOT what comes back — `r.timezone` is, because
* a run records the zone it was COMPUTED in and a definition's zone can be
* edited afterwards.
*/
const listInWindow = async ({ from, to, status = null, scope = null, seriesId = null, limit = 500 } = {}) => {
const where = ['r.scheduled_for >= ?', 'r.scheduled_for < ?']
const args = [from, to]
if (status) {
where.push('r.status = ?')
args.push(status)
}
if (scope !== null && scope !== undefined) {
where.push('r.scope = ?')
args.push(scope)
}
if (seriesId) {
where.push('d.series_id = ?')
args.push(seriesId)
}
const n = Math.min(Math.max(Number(limit) || 500, 1), 1000)
const rows = await query(
`SELECT r.*, d.title AS definition_title, d.slug AS definition_slug,
d.series_id AS series_id, se.name AS series_name, se.slug AS series_slug,
v.version AS version_number,
(SELECT COUNT(*) FROM event_run_steps s
WHERE s.run_id = r.id AND s.status = 'running' AND s.claim_expires_at IS NULL) AS waiting_steps
FROM event_runs r
JOIN event_definitions d ON d.id = r.definition_id
JOIN event_versions v ON v.id = r.version_id
LEFT JOIN event_series se ON se.id = d.series_id
WHERE ${where.join(' AND ')}
ORDER BY r.scheduled_for, r.id
LIMIT ${n}`,
args,
)
return rows.map(hydrate)
}
/**
* Point every not-yet-started occurrence of a definition at a new version.
*
* Publishing calls this, and the guard is the whole statement: `status =
* 'scheduled'` and `started_at IS NULL`. A run that has begun keeps the version
* it pinned, for ever, because that pin is what makes it explicable afterwards
* -- and a run that has NOT begun has nothing to explain yet.
*
* **Why re-pinning is the right answer and doing nothing is not** (org lead,
* 2026-09-02): occurrences are materialised a fortnight ahead, so on the day an
* editor fixes a typo there are already fourteen days of rows carrying the old
* spec. Left alone, the fix reaches none of them, and the operator's only
* recourse -- cancelling each one -- is worse: a cancelled row still holds its
* slot in `uq_evrun_occurrence`, so the occurrence does not come back on the new
* version, it disappears.
*
* Answers how many were moved, so publish can say so rather than leaving it to
* be noticed.
*/
const repinScheduled = async (definitionId, versionId) => {
const result = await query(
`UPDATE event_runs
SET version_id = ?
WHERE definition_id = ?
AND status = 'scheduled'
AND started_at IS NULL
AND version_id <> ?`,
[versionId, definitionId, versionId],
)
return Number(result?.affectedRows || 0)
}
/** The scheduled, not-yet-started occurrences a re-pin would move. */
const listScheduledFor = async (definitionId) =>
(
await query(
`SELECT id, version_id, scheduled_for FROM event_runs
WHERE definition_id = ? AND status = 'scheduled' AND started_at IS NULL
ORDER BY scheduled_for`,
[definitionId],
)
).map(hydrate)
/** The occurrence the unique key names, whether or not this call created it. */
const findOccurrence = async (definitionId, scope, scheduledFor) => {
const [row] = await query(
@@ -370,6 +456,9 @@ module.exports = {
list,
getById,
materialise,
listInWindow,
repinScheduled,
listScheduledFor,
findOccurrence,
countActiveForDefinition,
findDue,

View File

@@ -50,9 +50,21 @@ function renderConcurrencyKey(template, params) {
*
* `scheduledFor` defaults to now — "start now" is an occurrence whose instant is
* the present, not a separate concept, which is what keeps the runner's one
* materialise/advance path honest when Phase 4 adds recurrence on top.
* materialise/advance path honest now that Phase 4 has put recurrence on top.
*
* **Phase 4's expansion calls this, rather than a second insert path beside it.**
* That is deliberate: every check here — the definition is still `ready`, the
* version still has phases, the concurrency key renders, the first phase's steps
* are materialised with their idempotency keys — is one a scheduled occurrence
* needs at least as much as a hand-started one, because there is nobody watching
* when it happens. The `INSERT IGNORE` answering `created: false` is what makes
* it safe to call on every tick for every occurrence inside the horizon.
*/
async function create(definitionId, { scope = '', scheduledFor = null, rehearsal = false, params = null } = {}, userId) {
async function create(
definitionId,
{ scope = '', scheduledFor = null, rehearsal = false, params = null, source = 'manual' } = {},
userId,
) {
const definition = await definitionsDb.getById(definitionId)
if (!definition) return { ok: false, status: 404, errors: ['no such event definition'] }
if (definition.state !== 'ready') {
@@ -107,6 +119,11 @@ async function create(definitionId, { scope = '', scheduledFor = null, rehearsal
version: version.version,
scope: scopeValue,
rehearsal: Boolean(rehearsal),
// 'manual' is an admin pressing start; 'schedule' is the runner expanding
// a recurrence (Phase 4). Both produce the same row, and the log is the
// only place the difference is recorded — `started_by` is NULL for both a
// scheduled occurrence and one started by a since-deleted account.
source,
by: userId,
},
})

View File

@@ -1,17 +1,31 @@
// ── event_series — SQL only ────────────────────────────────────────────────
//
// EVENTS.md §D. The arc a definition may belong to. Phase 1 needs the reads —
// `event_definitions.series_id` is a foreign key and the definition save path
// has to check it resolves — and creating one is Phase 4's, where the calendar
// is what makes an arc visible.
// EVENTS.md §D. The arc a definition may belong to. Phase 1 needed only the
// reads — `event_definitions.series_id` is a foreign key and the definition save
// path has to check it resolves — and Phase 4 adds the writes, because the
// calendar is what makes an arc visible and a form cannot offer a value nobody
// can create.
//
// `ordering` here places a SERIES among the others on the calendar. A
// definition's place WITHIN its arc is `event_definitions.series_order`, which
// is the column an editor drags; the two are deliberately different columns on
// different tables and the schema comment says so.
const { query } = require('../../utils/db')
const list = async () =>
query('SELECT * FROM event_series ORDER BY ordering, name, id')
// `definition_count` is a correlated subquery rather than a join with a GROUP BY:
// the list is a handful of rows, and the delete path needs the same number to
// tell an operator what they are about to detach.
const SELECT_LIST = `
SELECT s.*,
(SELECT COUNT(*) FROM event_definitions d WHERE d.series_id = s.id) AS definition_count
FROM event_series s
`
const list = async () => query(`${SELECT_LIST} ORDER BY s.ordering, s.name, s.id`)
const getById = async (id) => {
const [row] = await query('SELECT * FROM event_series WHERE id = ?', [id])
const [row] = await query(`${SELECT_LIST} WHERE s.id = ?`, [id])
return row || null
}
@@ -20,4 +34,40 @@ const exists = async (id) => {
return Boolean(row)
}
module.exports = { list, getById, exists }
/** Does any OTHER series hold this slug? The uniqueness pre-check. */
const slugTaken = async (slug, exceptId = null) => {
const rows = exceptId
? await query('SELECT id FROM event_series WHERE slug = ? AND id <> ?', [slug, exceptId])
: await query('SELECT id FROM event_series WHERE slug = ?', [slug])
return rows.length > 0
}
const insert = async (s) => {
const result = await query(
`INSERT INTO event_series (name, slug, description, ordering, created_by)
VALUES (?, ?, ?, ?, ?)`,
[s.name, s.slug, s.description, s.ordering, s.created_by],
)
return Number(result.insertId)
}
const update = (id, s) =>
query(
`UPDATE event_series SET name = ?, slug = ?, description = ?, ordering = ? WHERE id = ?`,
[s.name, s.slug, s.description, s.ordering, id],
)
/**
* A hard delete, and the one place in this feature that is one.
*
* A series is a label rather than authored content: nothing pins one, no run
* references one, and `event_definitions.series_id` is `ON DELETE SET NULL`, so
* removing a series detaches its definitions and destroys nothing. That is why
* it is not archived the way a definition is — an archived label would be a
* state every calendar query has to remember for no benefit. The model answers
* with how many definitions were detached, so the operator learns what happened
* rather than discovering it on the calendar.
*/
const remove = (id) => query('DELETE FROM event_series WHERE id = ?', [id])
module.exports = { list, getById, exists, slugTaken, insert, update, remove }

View File

@@ -0,0 +1,93 @@
// ── Event series — the arc ─────────────────────────────────────────────────
//
// EVENTS.md §D and §I. "Royal Spy Mission → Risky Partner → Message From the
// Void" is continuity that exists nowhere in the tooling this feature replaces
// (§ "What the real calendar shows, and what it is missing": *no series or
// recurrence field*). One small table buys it, and this is the policy half.
//
// **Why the writes are `admin, editor` and not `admin`.** A series is authoring,
// and it is the same act as writing the definition that goes in it — §N2's
// narrow gate is about *committing the deployment to a run* (publish, start),
// which naming an arc does not do. An editor who can write the events but not
// the arc they belong to would have to ask an admin to type a title.
//
// **A slug is derived once and then frozen**, exactly as a definition's is: the
// public arc page lives at `/events/series/:slug` (Phase 14), and a slug that
// moved would break every link to it. Renaming the series is free.
const db = require('./eventSeries.db')
const { slugify, uniqueSlug } = require('../teams/teamSlug')
const MAX_NAME = 160
const MAX_DESCRIPTION = 2000
const trimOrNull = (v, max) => {
if (v === undefined || v === null) return null
const s = String(v).trim()
return s === '' ? null : s.slice(0, max)
}
const list = () => db.list()
const getById = (id) => db.getById(id)
async function validate(input, { existing = null } = {}) {
const errors = []
const body = input && typeof input === 'object' ? input : {}
const name = trimOrNull(body.name, MAX_NAME)
if (!name) errors.push('name is required')
const description = trimOrNull(body.description, MAX_DESCRIPTION)
const orderingRaw = body.ordering === undefined ? (existing?.ordering ?? 0) : body.ordering
const ordering = Number(orderingRaw)
if (!Number.isInteger(ordering) || ordering < 0 || ordering > 9999) {
errors.push('ordering must be an integer 0..9999')
}
if (errors.length) return { ok: false, errors }
return { ok: true, series: { name, description, ordering } }
}
async function create(input, userId) {
const checked = await validate(input)
if (!checked.ok) return { ok: false, status: 400, errors: checked.errors }
// The taken set is read here rather than inside `uniqueSlug` because that
// helper is pure — the same shape the team and definition paths use.
const taken = (await db.list()).map((s) => s.slug)
const slug = uniqueSlug(checked.series.name, taken, { fallback: 'series' })
const id = await db.insert({ ...checked.series, slug, created_by: userId || null })
return { ok: true, status: 201, series: await db.getById(id) }
}
async function update(id, input, userId) {
const existing = await db.getById(id)
if (!existing) return { ok: false, status: 404, errors: ['no such series'] }
const checked = await validate(input, { existing })
if (!checked.ok) return { ok: false, status: 400, errors: checked.errors }
// The slug is the existing one, deliberately: renaming a series must not move
// the address its arc page lives at.
await db.update(id, { ...checked.series, slug: existing.slug })
return { ok: true, status: 200, series: await db.getById(id) }
}
/**
* Delete a series, detaching whatever belonged to it.
*
* The count comes back so the caller can say *"3 events were detached"* rather
* than leaving an operator to notice on the calendar. `series_id` is
* `ON DELETE SET NULL`, so nothing is destroyed and re-attaching is a dropdown.
*/
async function remove(id) {
const existing = await db.getById(id)
if (!existing) return { ok: false, status: 404, errors: ['no such series'] }
await db.remove(id)
return { ok: true, status: 200, detached: Number(existing.definition_count || 0) }
}
module.exports = { list, getById, validate, create, update, remove, slugify, MAX_NAME }