chore: add duplicate organization pruning script [CM-766]

services/apps/script_executor_worker/src/activities.ts

@@ -2,10 +2,6 @@ import {
   blockMemberOrganizationAffiliation,
   getOrganizationMembers,
 } from './activities/block-organization-affiliation'
-import {
-  findDuplicateMembersAfterDate,
-  moveMemberActivityRelations,
-} from './activities/cleanup/duplicate-members'
 import { deleteMember, getMembersToCleanup, syncRemoveMember } from './activities/cleanup/member'
 import {
   deleteOrganization,
@@ -44,6 +40,10 @@ import {
   findMembersWithSameVerifiedEmailsInDifferentPlatforms,
 } from './activities/merge-members-with-similar-identities'
 import { getUnprocessedLLMApprovedSuggestions } from './activities/process-llm-verified-merges'
+import {
+  getOrganizationsToPrune,
+  pruneOrganization,
+} from './activities/prune-duplicate-organizations'
 import { deleteIndexedEntities } from './activities/sync/entity-index'
 import { getMembersForSync, syncMembersBatch } from './activities/sync/member'
 import { getOrganizationsForSync, syncOrganizationsBatch } from './activities/sync/organization'
@@ -78,12 +78,12 @@ export {
   deleteIndexedEntities,
   getUnprocessedLLMApprovedSuggestions,
   getWorkflowsCount,
-  findDuplicateMembersAfterDate,
-  moveMemberActivityRelations,
   getBotMembersWithOrgAffiliation,
   removeBotMemberOrganization,
   unlinkOrganizationFromBotActivities,
   blockMemberOrganizationAffiliation,
   getOrganizationMembers,
   calculateMemberAffiliations,
+  getOrganizationsToPrune,
+  pruneOrganization,
 }

services/apps/script_executor_worker/src/activities/prune-duplicate-organizations.ts

@@ -0,0 +1,25 @@
+import OrganizationRepository from '@crowd/data-access-layer/src/old/apps/script_executor_worker/organization.repo'
+
+import { svc } from '../main'
+
+export async function pruneOrganization(orgId: string): Promise<void> {
+  try {
+    const orgRepo = new OrganizationRepository(svc.postgres.writer.connection(), svc.log)
+    await orgRepo.pruneOrganization(orgId)
+  } catch (error) {
+    svc.log.error(error, 'Error pruning organization in database!')
+    throw error
+  }
+}
+
+export async function getOrganizationsToPrune(
+  batchSize: number,
+): Promise<{ id: string; displayName: string }[]> {
+  try {
+    const orgRepo = new OrganizationRepository(svc.postgres.reader.connection(), svc.log)
+    return orgRepo.getOrganizationsToPrune(batchSize)
+  } catch (error) {
+    svc.log.error(error, 'Error getting organizations to prune!')
+    throw error
+  }
+}

services/apps/script_executor_worker/src/types.ts

@@ -46,12 +46,6 @@ export interface IProcessLLMVerifiedMergesArgs extends IScriptBatchTestArgs {
   type: string
 }
 
-export interface ICleanupDuplicateMembersArgs extends IScriptBatchTestArgs {
-  cutoffDate?: string
-  checkByActivityIdentity?: boolean
-  checkByTwitterIdentity?: boolean
-}
-
 export interface IDedupActivityRelationsArgs extends IScriptBatchTestArgs {
   groupsPerRun?: number
   cursor?: Omit<IActivityRelationDuplicateGroup, 'activityIds'>

services/apps/script_executor_worker/src/workflows.ts

@@ -1,5 +1,4 @@
 import { blockOrganizationAffiliation } from './workflows/block-organization-affiliation'
-import { cleanupDuplicateMembers } from './workflows/cleanup/duplicate-members'
 import { cleanupMembers } from './workflows/cleanup/members'
 import { cleanupOrganizations } from './workflows/cleanup/organizations'
 import { dissectMember } from './workflows/dissectMember'
@@ -8,6 +7,7 @@ import { findAndMergeMembersWithSameVerifiedEmailsInDifferentPlatforms } from '.
 import { fixBotMembersAffiliation } from './workflows/fix-bot-members-affiliation'
 import { fixOrgIdentitiesWithWrongUrls } from './workflows/fixOrgIdentitiesWithWrongUrls'
 import { processLLMVerifiedMerges } from './workflows/processLLMVerifiedMerges'
+import { pruneDuplicateOrganizations } from './workflows/prune-duplicate-organizations'
 import { syncMembers } from './workflows/sync/members'
 import { syncOrganizations } from './workflows/sync/organizations'
 
@@ -21,7 +21,7 @@ export {
   cleanupMembers,
   cleanupOrganizations,
   processLLMVerifiedMerges,
-  cleanupDuplicateMembers,
+  pruneDuplicateOrganizations,
   fixBotMembersAffiliation,
   blockOrganizationAffiliation,
 }

services/apps/script_executor_worker/src/workflows/prune-duplicate-organizations.ts

@@ -0,0 +1,37 @@
+import { proxyActivities } from '@temporalio/workflow'
+
+import * as activities from '../activities'
+import { IScriptBatchTestArgs } from '../types'
+import { chunkArray } from '../utils/common'
+
+const { getOrganizationsToPrune, pruneOrganization, syncRemoveOrganization } = proxyActivities<
+  typeof activities
+>({
+  startToCloseTimeout: '30 minutes',
+  retry: { maximumAttempts: 3, backoffCoefficient: 3 },
+})
+
+export async function pruneDuplicateOrganizations(args: IScriptBatchTestArgs): Promise<void> {
+  const BATCH_SIZE = args.batchSize ?? 100
+
+  const organizationsToPrune = await getOrganizationsToPrune(BATCH_SIZE)
+
+  if (organizationsToPrune.length === 0) {
+    console.log('No more organizations to prune!')
+    return
+  }
+
+  const CHUNK_SIZE = 25
+
+  for (const chunk of chunkArray(organizationsToPrune, CHUNK_SIZE)) {
+    const cleanupTasks = chunk.map(async (o) => {
+      console.log('Pruning organization', o.displayName)
+      await syncRemoveOrganization(o.id)
+      return pruneOrganization(o.id)
+    })
+
+    await Promise.all(cleanupTasks).catch((err) => {
+      console.error('Error pruning organizations!', err)
+    })
+  }
+}

services/libs/data-access-layer/src/old/apps/script_executor_worker/organization.repo.ts

@@ -179,6 +179,87 @@ class OrganizationRepository {
       { organizationId, limit, offset },
     )
   }
+
+  public async pruneOrganization(organizationId: string): Promise<void> {
+    const tablesToDelete = [
+      { name: 'organizationNoMerge', conditions: ['organizationId', 'noMergeId'] },
+      { name: 'organizationToMerge', conditions: ['organizationId', 'toMergeId'] },
+      { name: 'organizationToMergeRaw', conditions: ['organizationId', 'toMergeId'] },
+      { name: 'organizationEnrichmentCache', conditions: ['organizationId'] },
+      { name: 'organizationEnrichments', conditions: ['organizationId'] },
+      { name: 'memberSegmentAffiliations', conditions: ['organizationId'] },
+      { name: 'organizationSegmentsAgg', conditions: ['organizationId'] },
+      { name: 'organizationSegments', conditions: ['organizationId'] },
+      { name: 'orgAttributes', conditions: ['organizationId'] },
+      { name: 'organizationIdentities', conditions: ['organizationId'] },
+      { name: 'memberOrganizations', conditions: ['organizationId'] },
+      { name: 'organizations', conditions: ['id'] },
+    ]
+
+    await this.connection.tx(async (tx) => {
+      for (const table of tablesToDelete) {
+        const whereClause = table.conditions
+          .map((field) => `"${field}" = $(organizationId)`)
+          .join(' OR ')
+        await tx.none(`DELETE FROM "${table.name}" WHERE ${whereClause}`, { organizationId })
+      }
+    })
+  }
+
+  public async getOrganizationsToPrune(
+    batchSize: number,
+  ): Promise<{ id: string; displayName: string }[]> {
+    return this.connection.query(
+      `
+      WITH email_providers AS (
+        SELECT unnest(ARRAY[
+            'gmail.com', 'gmail.co.uk', 'gmail.com.au', 'gmail.com.tr',
+            'yahoo.com', 'yahoo.co.uk', 'yahoo.com.br', 'yahoo.co.in', 'yahoo.fr', 'yahoo.es', 'yahoo.it', 'yahoo.de', 'yahoo.ca', 'yahoo.com.au', 'yahoo.in', 'yahoo.co.jp', 'yahoo.com.ar', 'yahoo.com.mx', 'yahoo.co.id', 'yahoo.com.sg', 'yahoo.co.za', 'yahoo.com.ph', 'yahoo.com.tw', 'yahoo.com.hk', 'yahoo.com.vn',
+            'hotmail.com', 'hotmail.co.uk', 'hotmail.fr', 'hotmail.ca', 'hotmail.it', 'hotmail.es', 'hotmail.de', 'hotmail.com.au', 'hotmail.com.mx',
+            'icloud.com', 'icloud.com.cn',
+            'fastmail.com', 'tutanota.com', 'tuta.io',
+            'gmx.com', 'gmx.de', 'gmx.net', 'gmx.at', 'gmx.ch', 'gmx.fr', 'gmx.co.uk',
+            'aol.com', 'aol.co.uk', 'aol.fr', 'aol.de',
+            'msn.com', 'wanadoo.fr', 'orange.fr', 'comcast.net',
+            'live.com', 'live.co.uk', 'live.fr', 'live.nl', 'live.it', 'live.com.au', 'live.ca', 'live.cn',
+            'rediffmail.com', 'sify.com', 'indiatimes.com', 'free.fr', 'web.de',
+            'yandex.ru', 'yandex.com', 'yandex.com.tr', 'ya.ru',
+            'ymail.com', 'libero.it',
+            'outlook.com', 'outlook.fr', 'outlook.co.uk', 'outlook.de', 'outlook.es', 'outlook.it', 'outlook.com.au', 'outlook.com.br', 'outlook.com.mx', 'outlook.co.jp', 'outlook.in', 'outlook.com.sg', 'outlook.co.za', 'outlook.co.in',
+            'uol.com.br', 'bol.com.br',
+            'mail.ru', 'inbox.ru', 'list.ru', 'bk.ru',
+            'mail.com', 'mail.de', 'mail.co.uk',
+            'cox.net', 'sbcglobal.net', 'sfr.fr', 'verizon.net', 'googlemail.com', 'ig.com.br', 'bigpond.com', 'bigpond.net.au', 'terra.com.br', 'neuf.fr', 'alice.it', 'rocketmail.com', 'att.net', 'laposte.net', 'bellsouth.net', 'charter.net', 'rambler.ru', 'tiscali.it', 'tiscali.co.uk', 'shaw.ca', 'sky.com', 'earthlink.net', 'optonline.net', 'freenet.de', 't-online.de', 'aliceadsl.fr', 'virgilio.it', 'home.nl', 'qq.com', 'vip.qq.com', 'telenet.be', 'pandora.be', 'me.com', 'voila.fr', 'planet.nl', 'tin.it', 'ntlworld.com', 'arcor.de', 'frontiernet.net', 'hetnet.nl', 'zonnet.nl', 'club-internet.fr', 'juno.com', 'optusnet.com.au', 'blueyonder.co.uk', 'bluewin.ch', 'skynet.be', 'sympatico.ca', 'windstream.net', 'mac.com', 'centurytel.net', 'chello.nl', 'aim.com',
+            'protonmail.com', 'protonmail.ch', 'proton.me', 'pm.me', 'duck.com',
+            'zoho.com', 'zohomail.com',
+            'users.noreply.github.com',
+            '126.com', '139.com', '163.com', '188.com', 'foxmail.com', 'tom.com', '21cn.com', 'yeah.net',
+            'naver.com', 'daum.net', 'hanmail.net',
+            'hey.com', 'inbox.com', 'lycos.com', 'excite.com', 'hushmail.com', 'mailfence.com', 'mailbox.org', 'posteo.de', 'startmail.com', 'runbox.com', 'countermail.com', 'mynet.com',
+            'wp.pl', 'onet.pl', 'interia.pl', 'o2.pl',
+            'seznam.cz', 'centrum.cz',
+            'mailinator.com', 'guerrillamail.com', '10minutemail.com', 'tempmail.com'
+        ]) AS provider
+      )
+      SELECT DISTINCT o.id, o."displayName"
+      FROM organizations o
+      INNER JOIN "organizationIdentities" oi
+          ON o.id = oi."organizationId"
+      INNER JOIN email_providers ep
+          ON LOWER(oi.value) = ep.provider
+      WHERE o."deletedAt" IS NULL
+        AND oi.type = 'primary-domain'
+      AND NOT EXISTS (
+          SELECT 1
+          FROM "memberOrganizations" mo
+          WHERE mo."organizationId" = o.id
+            AND (mo.title IS NOT NULL AND mo.title != '')
+            AND (mo.source IS NOT NULL AND mo.source NOT IN ('email-domain'))
+      )
+      `,
+      { batchSize },
+    )
+  }
 }
 
 export default OrganizationRepository