server/lib/activitypub/crawl.ts

   1 import * as Bluebird from 'bluebird'
   2 import { URL } from 'url'
   3 import { ActivityPubOrderedCollection } from '../../../shared/models/activitypub'
   4 import { logger } from '../../helpers/logger'
   5 import { doJSONRequest } from '../../helpers/requests'
   6 import { ACTIVITY_PUB, REQUEST_TIMEOUT, WEBSERVER } from '../../initializers/constants'
   7
   8 type HandlerFunction<T> = (items: T[]) => (Promise<any> | Bluebird<any>)
   9 type CleanerFunction = (startedDate: Date) => (Promise<any> | Bluebird<any>)
  10
  11 async function crawlCollectionPage <T> (argUrl: string, handler: HandlerFunction<T>, cleaner?: CleanerFunction) {
  12   let url = argUrl
  13
  14   logger.info('Crawling ActivityPub data on %s.', url)
  15
  16   const options = {
  17     activityPub: true,
  18     timeout: REQUEST_TIMEOUT
  19   }
  20
  21   const startDate = new Date()
  22
  23   const response = await doJSONRequest<ActivityPubOrderedCollection<T>>(url, options)
  24   const firstBody = response.body
  25
  26   const limit = ACTIVITY_PUB.FETCH_PAGE_LIMIT
  27   let i = 0
  28   let nextLink = firstBody.first
  29   while (nextLink && i < limit) {
  30     let body: any
  31
  32     if (typeof nextLink === 'string') {
  33       // Don't crawl ourselves
  34       const remoteHost = new URL(nextLink).host
  35       if (remoteHost === WEBSERVER.HOST) continue
  36
  37       url = nextLink
  38
  39       const res = await doJSONRequest<ActivityPubOrderedCollection<T>>(url, options)
  40       body = res.body
  41     } else {
  42       // nextLink is already the object we want
  43       body = nextLink
  44     }
  45
  46     nextLink = body.next
  47     i++
  48
  49     if (Array.isArray(body.orderedItems)) {
  50       const items = body.orderedItems
  51       logger.info('Processing %i ActivityPub items for %s.', items.length, url)
  52
  53       await handler(items)
  54     }
  55   }
  56
  57   if (cleaner) await cleaner(startDate)
  58 }
  59
  60 export {
  61   crawlCollectionPage
  62 }