PeerTube_original/server/middlewares/cache/shared/api-cache.ts
Alecks Gates cb0eda5602
Add Podcast RSS feeds (#5487)
* Initial test implementation of Podcast RSS

This is a pretty simple implementation to add support for The Podcast Namespace in RSS -- instead of affecting the existing RSS implementation, this adds a new UI option.

I attempted to retain compatibility with the rest of the RSS feed implementation as much as possible and have created a temporary fork of the "pfeed" library to support this effort.

* Update to pfeed-podcast 1.2.2

* Initial test implementation of Podcast RSS

This is a pretty simple implementation to add support for The Podcast Namespace in RSS -- instead of affecting the existing RSS implementation, this adds a new UI option.

I attempted to retain compatibility with the rest of the RSS feed implementation as much as possible and have created a temporary fork of the "pfeed" library to support this effort.

* Update to pfeed-podcast 1.2.2

* Initial test implementation of Podcast RSS

This is a pretty simple implementation to add support for The Podcast Namespace in RSS -- instead of affecting the existing RSS implementation, this adds a new UI option.

I attempted to retain compatibility with the rest of the RSS feed implementation as much as possible and have created a temporary fork of the "pfeed" library to support this effort.

* Update to pfeed-podcast 1.2.2

* Add correct feed image to RSS channel

* Prefer HLS videos for podcast RSS

Remove video/stream titles, add optional height attribute to podcast RSS

* Prefix podcast RSS images with root server URL

* Add optional video query support to include captions

* Add transcripts & person images to podcast RSS feed

* Prefer webseed/webtorrent files over HLS fragmented mp4s

* Experimentally adding podcast fields to basic config page

* Add validation for new basic config fields

* Don't include "content" in podcast feed, use full description for "description"

* Initial test implementation of Podcast RSS

This is a pretty simple implementation to add support for The Podcast Namespace in RSS -- instead of affecting the existing RSS implementation, this adds a new UI option.

I attempted to retain compatibility with the rest of the RSS feed implementation as much as possible and have created a temporary fork of the "pfeed" library to support this effort.

* Update to pfeed-podcast 1.2.2

* Add correct feed image to RSS channel

* Prefer HLS videos for podcast RSS

Remove video/stream titles, add optional height attribute to podcast RSS

* Prefix podcast RSS images with root server URL

* Add optional video query support to include captions

* Add transcripts & person images to podcast RSS feed

* Prefer webseed/webtorrent files over HLS fragmented mp4s

* Experimentally adding podcast fields to basic config page

* Add validation for new basic config fields

* Don't include "content" in podcast feed, use full description for "description"

* Add medium/socialInteract to podcast RSS feeds. Use HTML for description

* Change base production image to bullseye, install prosody in image

* Add liveItem and trackers to Podcast RSS feeds

Remove height from alternateEnclosure, replaced with title.

* Clear Podcast RSS feed cache when live streams start/end

* Upgrade to Node 16

* Refactor clearCacheRoute to use ApiCache

* Remove unnecessary type hint

* Update dockerfile to node 16, install python-is-python2

* Use new file paths for captions/playlists

* Fix legacy videos in RSS after migration to object storage

* Improve method of identifying non-fragmented mp4s in podcast RSS feeds

* Don't include fragmented MP4s in podcast RSS feeds

* Add experimental support for podcast:categories on the podcast RSS item

* Fix undefined category when no videos exist

Allows for empty feeds to exist (important for feeds that might only go live)

* Add support for podcast:locked -- user has to opt in to show their email

* Use comma for podcast:categories delimiter

* Make cache clearing async

* Fix merge, temporarily test with pfeed-podcast

* Syntax changes

* Add EXT_MIMETYPE constants for captions

* Update & fix tests, fix enclosure mimetypes, remove admin email

* Add test for podacst:socialInteract

* Add filters hooks for podcast customTags

* Remove showdown, updated to pfeed-podcast 6.1.2

* Add 'action:api.live-video.state.updated' hook

* Avoid assigning undefined category to podcast feeds

* Remove nvmrc

* Remove comment

* Remove unused podcast config

* Remove more unused podcast config

* Fix MChannelAccountDefault type hint missed in merge

* Remove extra line

* Re-add newline in config

* Fix lint errors for isEmailPublic

* Fix thumbnails in podcast feeds

* Requested changes based on review

* Provide podcast rss 2.0 only on video channels

* Misc cleanup for a less messy PR

* Lint fixes

* Remove pfeed-podcast

* Add peertube version to new hooks

* Don't use query include, remove TODO

* Remove film medium hack

* Clear podcast rss cache before video/channel update hooks

* Clear podcast rss cache before video uploaded/deleted hooks

* Refactor podcast feed cache clearing

* Set correct person name from video channel

* Styling

* Fix tests

---------

Co-authored-by: Chocobozzz <me@florianbigard.com>
2023-05-22 16:00:05 +02:00

311 lines
8.5 KiB
TypeScript

// Thanks: https://github.com/kwhitley/apicache
// We duplicated the library because it is unmaintened and prevent us to upgrade to recent NodeJS versions
import express from 'express'
import { OutgoingHttpHeaders } from 'http'
import { isTestInstance, parseDurationToMs } from '@server/helpers/core-utils'
import { logger } from '@server/helpers/logger'
import { Redis } from '@server/lib/redis'
import { asyncMiddleware } from '@server/middlewares'
import { HttpStatusCode } from '@shared/models'
export interface APICacheOptions {
headerBlacklist?: string[]
excludeStatus?: HttpStatusCode[]
}
interface CacheObject {
status: number
headers: OutgoingHttpHeaders
data: any
encoding: BufferEncoding
timestamp: number
}
export class ApiCache {
private readonly options: APICacheOptions
private readonly timers: { [ id: string ]: NodeJS.Timeout } = {}
private readonly index = {
groups: [] as string[],
all: [] as string[]
}
// Cache keys per group
private groups: { [groupIndex: string]: string[] } = {}
constructor (options: APICacheOptions) {
this.options = {
headerBlacklist: [],
excludeStatus: [],
...options
}
}
buildMiddleware (strDuration: string) {
const duration = parseDurationToMs(strDuration)
return asyncMiddleware(
async (req: express.Request, res: express.Response, next: express.NextFunction) => {
const key = this.getCacheKey(req)
const redis = Redis.Instance.getClient()
if (!Redis.Instance.isConnected()) return this.makeResponseCacheable(res, next, key, duration)
try {
const obj = await redis.hgetall(key)
if (obj?.response) {
return this.sendCachedResponse(req, res, JSON.parse(obj.response), duration)
}
return this.makeResponseCacheable(res, next, key, duration)
} catch (err) {
return this.makeResponseCacheable(res, next, key, duration)
}
}
)
}
clearGroupSafe (group: string) {
const run = async () => {
const cacheKeys = this.groups[group]
if (!cacheKeys) return
for (const key of cacheKeys) {
try {
await this.clear(key)
} catch (err) {
logger.error('Cannot clear ' + key, { err })
}
}
delete this.groups[group]
}
void run()
}
private getCacheKey (req: express.Request) {
return Redis.Instance.getPrefix() + 'api-cache-' + req.originalUrl
}
private shouldCacheResponse (response: express.Response) {
if (!response) return false
if (this.options.excludeStatus.includes(response.statusCode)) return false
return true
}
private addIndexEntries (key: string, res: express.Response) {
this.index.all.unshift(key)
const groups = res.locals.apicacheGroups || []
for (const group of groups) {
if (!this.groups[group]) this.groups[group] = []
this.groups[group].push(key)
}
}
private filterBlacklistedHeaders (headers: OutgoingHttpHeaders) {
return Object.keys(headers)
.filter(key => !this.options.headerBlacklist.includes(key))
.reduce((acc, header) => {
acc[header] = headers[header]
return acc
}, {})
}
private createCacheObject (status: number, headers: OutgoingHttpHeaders, data: any, encoding: BufferEncoding) {
return {
status,
headers: this.filterBlacklistedHeaders(headers),
data,
encoding,
// Seconds since epoch, used to properly decrement max-age headers in cached responses.
timestamp: new Date().getTime() / 1000
} as CacheObject
}
private async cacheResponse (key: string, value: object, duration: number) {
const redis = Redis.Instance.getClient()
if (Redis.Instance.isConnected()) {
await Promise.all([
redis.hset(key, 'response', JSON.stringify(value)),
redis.hset(key, 'duration', duration + ''),
redis.expire(key, duration / 1000)
])
}
// add automatic cache clearing from duration, includes max limit on setTimeout
this.timers[key] = setTimeout(() => {
this.clear(key)
.catch(err => logger.error('Cannot clear Redis key %s.', key, { err }))
}, Math.min(duration, 2147483647))
}
private accumulateContent (res: express.Response, content: any) {
if (!content) return
if (typeof content === 'string') {
res.locals.apicache.content = (res.locals.apicache.content || '') + content
return
}
if (Buffer.isBuffer(content)) {
let oldContent = res.locals.apicache.content
if (typeof oldContent === 'string') {
oldContent = Buffer.from(oldContent)
}
if (!oldContent) {
oldContent = Buffer.alloc(0)
}
res.locals.apicache.content = Buffer.concat(
[ oldContent, content ],
oldContent.length + content.length
)
return
}
res.locals.apicache.content = content
}
private makeResponseCacheable (res: express.Response, next: express.NextFunction, key: string, duration: number) {
const self = this
res.locals.apicache = {
write: res.write,
writeHead: res.writeHead,
end: res.end,
cacheable: true,
content: undefined,
headers: undefined
}
// Patch express
res.writeHead = function () {
if (self.shouldCacheResponse(res)) {
res.setHeader('cache-control', 'max-age=' + (duration / 1000).toFixed(0))
} else {
res.setHeader('cache-control', 'no-cache, no-store, must-revalidate')
}
res.locals.apicache.headers = Object.assign({}, res.getHeaders())
return res.locals.apicache.writeHead.apply(this, arguments as any)
}
res.write = function (chunk: any) {
self.accumulateContent(res, chunk)
return res.locals.apicache.write.apply(this, arguments as any)
}
res.end = function (content: any, encoding: BufferEncoding) {
if (self.shouldCacheResponse(res)) {
self.accumulateContent(res, content)
if (res.locals.apicache.cacheable && res.locals.apicache.content) {
self.addIndexEntries(key, res)
const headers = res.locals.apicache.headers || res.getHeaders()
const cacheObject = self.createCacheObject(
res.statusCode,
headers,
res.locals.apicache.content,
encoding
)
self.cacheResponse(key, cacheObject, duration)
.catch(err => logger.error('Cannot cache response', { err }))
}
}
res.locals.apicache.end.apply(this, arguments as any)
} as any
next()
}
private sendCachedResponse (request: express.Request, response: express.Response, cacheObject: CacheObject, duration: number) {
const headers = response.getHeaders()
if (isTestInstance()) {
Object.assign(headers, {
'x-api-cache-cached': 'true'
})
}
Object.assign(headers, this.filterBlacklistedHeaders(cacheObject.headers || {}), {
// Set properly decremented max-age header
// This ensures that max-age is in sync with the cache expiration
'cache-control':
'max-age=' +
Math.max(
0,
(duration / 1000 - (new Date().getTime() / 1000 - cacheObject.timestamp))
).toFixed(0)
})
// unstringify buffers
let data = cacheObject.data
if (data && data.type === 'Buffer') {
data = typeof data.data === 'number'
? Buffer.alloc(data.data)
: Buffer.from(data.data)
}
// Test Etag against If-None-Match for 304
const cachedEtag = cacheObject.headers.etag
const requestEtag = request.headers['if-none-match']
if (requestEtag && cachedEtag === requestEtag) {
response.writeHead(304, headers)
return response.end()
}
response.writeHead(cacheObject.status || 200, headers)
return response.end(data, cacheObject.encoding)
}
private async clear (target: string) {
const redis = Redis.Instance.getClient()
if (target) {
clearTimeout(this.timers[target])
delete this.timers[target]
try {
await redis.del(target)
} catch (err) {
logger.error('Cannot delete %s in redis cache.', target, { err })
}
this.index.all = this.index.all.filter(key => key !== target)
} else {
for (const key of this.index.all) {
clearTimeout(this.timers[key])
delete this.timers[key]
try {
await redis.del(key)
} catch (err) {
logger.error('Cannot delete %s in redis cache.', key, { err })
}
}
this.index.all = []
}
return this.index
}
}