From 92f6dd96c0ba234e891043dd76268469dcf1ec6e Mon Sep 17 00:00:00 2001 From: Nolan Lawson Date: Tue, 13 Feb 2018 19:34:37 -0800 Subject: [PATCH] implement periodic db cleanup --- routes/_database/accountsAndRelationships.js | 6 +- routes/_database/cleanup.js | 135 +++++++++++++++++++ routes/_database/constants.js | 5 + routes/_database/databaseLifecycle.js | 37 ++--- routes/_database/helpers.js | 29 ++++ routes/_database/timelines.js | 43 ++---- 6 files changed, 202 insertions(+), 53 deletions(-) create mode 100644 routes/_database/cleanup.js diff --git a/routes/_database/accountsAndRelationships.js b/routes/_database/accountsAndRelationships.js index b01e032..e61a713 100644 --- a/routes/_database/accountsAndRelationships.js +++ b/routes/_database/accountsAndRelationships.js @@ -1,13 +1,13 @@ import { ACCOUNTS_STORE, RELATIONSHIPS_STORE } from './constants' import { accountsCache, relationshipsCache } from './cache' -import { getGenericEntityWithId, setGenericEntityWithId } from './helpers' +import { cloneForStorage, getGenericEntityWithId, setGenericEntityWithId } from './helpers' export async function getAccount (instanceName, accountId) { return getGenericEntityWithId(ACCOUNTS_STORE, accountsCache, instanceName, accountId) } export async function setAccount (instanceName, account) { - return setGenericEntityWithId(ACCOUNTS_STORE, accountsCache, instanceName, account) + return setGenericEntityWithId(ACCOUNTS_STORE, accountsCache, instanceName, cloneForStorage(account)) } export async function getRelationship (instanceName, accountId) { @@ -15,5 +15,5 @@ export async function getRelationship (instanceName, accountId) { } export async function setRelationship (instanceName, relationship) { - return setGenericEntityWithId(RELATIONSHIPS_STORE, relationshipsCache, instanceName, relationship) + return setGenericEntityWithId(RELATIONSHIPS_STORE, relationshipsCache, instanceName, cloneForStorage(relationship)) } diff --git a/routes/_database/cleanup.js b/routes/_database/cleanup.js new file mode 100644 index 0000000..ce84f09 --- /dev/null +++ b/routes/_database/cleanup.js @@ -0,0 +1,135 @@ +import { dbPromise, getDatabase } from './databaseLifecycle' +import { scheduleIdleTask } from '../_utils/scheduleIdleTask' +import { + ACCOUNTS_STORE, + NOTIFICATION_TIMELINES_STORE, + NOTIFICATIONS_STORE, + RELATIONSHIPS_STORE, + STATUS_TIMELINES_STORE, + STATUSES_STORE, + TIMESTAMP +} from './constants' +import debounce from 'lodash/debounce' +import { store } from '../_store/store' +import { mark, stop } from '../_utils/marks' + +const BATCH_SIZE = 20 +const TIME_AGO = 14 * 24 * 60 * 60 * 1000 // two weeks ago +const DELAY = 5 * 60 * 1000 // five minutes + +function batchedGetAll(callGetAll, callback) { + function nextBatch() { + callGetAll().onsuccess = function (e) { + let results = e.target.result + callback(results) + if (results.length) { + nextBatch() + } + } + } + nextBatch() +} + +function cleanupStatuses(statusesStore, statusTimelinesStore, cutoff) { + batchedGetAll( + () => statusesStore.index(TIMESTAMP).getAll(IDBKeyRange.upperBound(cutoff), BATCH_SIZE), + results => { + results.forEach(result => { + statusesStore.delete(result.id) + let req = statusTimelinesStore.index('statusId').getAll(IDBKeyRange.only(result.id)) + req.onsuccess = e => { + let results = e.target.result + results.forEach(result => { + statusTimelinesStore.delete(result.id) + }) + } + }) + } + ) +} + +function cleanupNotifications(notificationsStore, notificationTimelinesStore, cutoff) { + batchedGetAll( + () => notificationsStore.index(TIMESTAMP).getAll(IDBKeyRange.upperBound(cutoff), BATCH_SIZE), + results => { + results.forEach(result => { + notificationsStore.delete(result.id) + let req = notificationTimelinesStore.index('notificationId').getAll(IDBKeyRange.only(result.id)) + req.onsuccess = e => { + let results = e.target.result + results.forEach(result => { + notificationTimelinesStore.delete(result.id) + }) + } + }) + } + ) +} + +function cleanupAccounts(accountsStore, cutoff) { + batchedGetAll( + () => accountsStore.index(TIMESTAMP).getAll(IDBKeyRange.upperBound(cutoff), BATCH_SIZE), + (results) => { + results.forEach(result => { + accountsStore.delete(result.id) + }) + } + ) +} + +function cleanupRelationships(relationshipsStore, cutoff) { + batchedGetAll( + () => relationshipsStore.index(TIMESTAMP).getAll(IDBKeyRange.upperBound(cutoff), BATCH_SIZE), + (results) => { + results.forEach(result => { + relationshipsStore.delete(result.id) + }) + } + ) +} + +async function cleanup(instanceName) { + console.log('cleanup', instanceName) + mark(`cleanup:${instanceName}`) + let db = await getDatabase(instanceName) + let storeNames = [ + STATUSES_STORE, + STATUS_TIMELINES_STORE, + NOTIFICATIONS_STORE, + NOTIFICATION_TIMELINES_STORE, + ACCOUNTS_STORE, + RELATIONSHIPS_STORE + ] + await dbPromise(db, storeNames, 'readwrite', (stores) => { + let [ + statusesStore, + statusTimelinesStore, + notificationsStore, + notificationTimelinesStore, + accountsStore, + relationshipsStore + ] = stores + + let cutoff = Date.now() - TIME_AGO + + cleanupStatuses(statusesStore, statusTimelinesStore, cutoff) + cleanupNotifications(notificationsStore, notificationTimelinesStore, cutoff) + cleanupAccounts(accountsStore, cutoff) + cleanupRelationships(relationshipsStore, cutoff) + }) + stop(`cleanup:${instanceName}`) +} + +function doCleanup(instanceName) { + scheduleIdleTask(() => cleanup(instanceName)) +} + +function scheduledCleanup () { + console.log('scheduledCleanup') + let instances = store.get('loggedInInstancesInOrder') + for (let instance of instances) { + doCleanup(instance) + } +} + +export const scheduleCleanup = debounce(() => scheduleIdleTask(scheduledCleanup), DELAY) \ No newline at end of file diff --git a/routes/_database/constants.js b/routes/_database/constants.js index e69def9..59c2a30 100644 --- a/routes/_database/constants.js +++ b/routes/_database/constants.js @@ -6,3 +6,8 @@ export const RELATIONSHIPS_STORE = 'relationships' export const NOTIFICATIONS_STORE = 'notifications' export const NOTIFICATION_TIMELINES_STORE = 'notification_timelines' export const PINNED_STATUSES_STORE = 'pinned_statuses' + +export const TIMESTAMP = '__pinafore_ts' +export const ACCOUNT_ID = '__pinafore_acct_id' +export const STATUS_ID = '__pinafore_status_id' +export const REBLOG_ID = '__pinafore_reblog_id' \ No newline at end of file diff --git a/routes/_database/databaseLifecycle.js b/routes/_database/databaseLifecycle.js index 78ff571..a51e9f5 100644 --- a/routes/_database/databaseLifecycle.js +++ b/routes/_database/databaseLifecycle.js @@ -6,13 +6,18 @@ import { RELATIONSHIPS_STORE, NOTIFICATIONS_STORE, NOTIFICATION_TIMELINES_STORE, - PINNED_STATUSES_STORE + PINNED_STATUSES_STORE, + TIMESTAMP } from './constants' const openReqs = {} const databaseCache = {} -const DB_VERSION = 2 +const DB_VERSION = 1 + +function objectStore(db, name, keyPath) { + return db.createObjectStore(name, {keyPath: keypPath}) +} export function getDatabase (instanceName) { if (!instanceName) { @@ -31,20 +36,20 @@ export function getDatabase (instanceName) { } req.onupgradeneeded = (e) => { let db = req.result - if (e.oldVersion < 1) { - db.createObjectStore(META_STORE, {keyPath: 'key'}) - db.createObjectStore(STATUSES_STORE, {keyPath: 'id'}) - db.createObjectStore(ACCOUNTS_STORE, {keyPath: 'id'}) - db.createObjectStore(RELATIONSHIPS_STORE, {keyPath: 'id'}) - db.createObjectStore(NOTIFICATIONS_STORE, {keyPath: 'id'}) - db.createObjectStore(STATUS_TIMELINES_STORE, {keyPath: 'id'}) - .createIndex('statusId', 'statusId') - db.createObjectStore(NOTIFICATION_TIMELINES_STORE, {keyPath: 'id'}) - .createIndex('notificationId', 'notificationId') - } - if (e.oldVersion < 2) { - db.createObjectStore(PINNED_STATUSES_STORE, {keyPath: 'id'}) - } + db.createObjectStore(STATUSES_STORE, {keyPath: 'id'}) + .createIndex(TIMESTAMP, TIMESTAMP) + db.createObjectStore(STATUS_TIMELINES_STORE, {keyPath: 'id'}) + .createIndex('statusId', 'statusId') + db.createObjectStore(NOTIFICATIONS_STORE, {keyPath: 'id'}) + .createIndex(TIMESTAMP, TIMESTAMP) + db.createObjectStore(NOTIFICATION_TIMELINES_STORE, {keyPath: 'id'}) + .createIndex('notificationId', 'notificationId') + db.createObjectStore(ACCOUNTS_STORE, {keyPath: 'id'}) + .createIndex(TIMESTAMP, TIMESTAMP) + db.createObjectStore(RELATIONSHIPS_STORE, {keyPath: 'id'}) + .createIndex(TIMESTAMP, TIMESTAMP) + db.createObjectStore(META_STORE, {keyPath: 'key'}) + db.createObjectStore(PINNED_STATUSES_STORE, {keyPath: 'id'}) } req.onsuccess = () => resolve(req.result) }) diff --git a/routes/_database/helpers.js b/routes/_database/helpers.js index e1c8d24..6d951ac 100644 --- a/routes/_database/helpers.js +++ b/routes/_database/helpers.js @@ -1,5 +1,6 @@ import { dbPromise, getDatabase } from './databaseLifecycle' import { getInCache, hasInCache, setInCache } from './cache' +import { ACCOUNT_ID, REBLOG_ID, STATUS_ID, TIMESTAMP } from './constants' export async function getGenericEntityWithId (store, cache, instanceName, id) { if (hasInCache(cache, instanceName, id)) { @@ -20,3 +21,31 @@ export async function setGenericEntityWithId (store, cache, instanceName, entity store.put(entity) }) } + +export function cloneForStorage (obj) { + let res = {} + let keys = Object.keys(obj) + for (let key of keys) { + let value = obj[key] + // save storage space by skipping nulls, 0s, falses, empty strings, and empty arrays + if (!value || (Array.isArray(value) && value.length === 0)) { + continue + } + switch (key) { + case 'account': + res[ACCOUNT_ID] = value.id + break + case 'status': + res[STATUS_ID] = value.id + break + case 'reblog': + res[REBLOG_ID] = value.id + break + default: + res[key] = value + break + } + } + res[TIMESTAMP] = Date.now() + return res +} \ No newline at end of file diff --git a/routes/_database/timelines.js b/routes/_database/timelines.js index 46f4daf..489bd8d 100644 --- a/routes/_database/timelines.js +++ b/routes/_database/timelines.js @@ -1,18 +1,19 @@ import { toPaddedBigInt, toReversePaddedBigInt } from './utils' +import { cloneForStorage } from './helpers' import { dbPromise, getDatabase } from './databaseLifecycle' import { accountsCache, getInCache, hasInCache, notificationsCache, setInCache, statusesCache } from './cache' +import { scheduleCleanup } from './cleanup' import { ACCOUNTS_STORE, NOTIFICATION_TIMELINES_STORE, NOTIFICATIONS_STORE, PINNED_STATUSES_STORE, STATUS_TIMELINES_STORE, - STATUSES_STORE + STATUSES_STORE, + ACCOUNT_ID, + REBLOG_ID, + STATUS_ID } from './constants' - -const TIMESTAMP = '__pinafore_ts' -const ACCOUNT_ID = '__pinafore_acct_id' -const STATUS_ID = '__pinafore_status_id' -const REBLOG_ID = '__pinafore_reblog_id' +import { scheduleCleanup } from './cleanup' function createTimelineKeyRange (timeline, maxId) { let negBigInt = maxId && toReversePaddedBigInt(maxId) @@ -29,34 +30,6 @@ function createKeyRangeForStatusThread (timeline) { return IDBKeyRange.bound(start, end, true, true) } -function cloneForStorage (obj) { - let res = {} - let keys = Object.keys(obj) - for (let key of keys) { - let value = obj[key] - // save storage space by skipping nulls, 0s, falses, empty strings, and empty arrays - if (!value || (Array.isArray(value) && value.length === 0)) { - continue - } - switch (key) { - case 'account': - res[ACCOUNT_ID] = value.id - break - case 'status': - res[STATUS_ID] = value.id - break - case 'reblog': - res[REBLOG_ID] = value.id - break - default: - res[key] = value - break - } - } - res[TIMESTAMP] = Date.now() - return res -} - function cacheStatus (status, instanceName) { setInCache(statusesCache, instanceName, status.id, status) setInCache(accountsCache, instanceName, status.account.id, status.account) @@ -202,6 +175,7 @@ function createTimelineId (timeline, id) { } async function insertTimelineNotifications (instanceName, timeline, notifications) { + /* no await */ scheduleCleanup() for (let notification of notifications) { setInCache(notificationsCache, instanceName, notification.id, notification) setInCache(accountsCache, instanceName, notification.account.id, notification.account) @@ -221,6 +195,7 @@ async function insertTimelineNotifications (instanceName, timeline, notification } async function insertTimelineStatuses (instanceName, timeline, statuses) { + /* no await */ scheduleCleanup() for (let status of statuses) { cacheStatus(status, instanceName) }