Auto complete: a task to build user contacts from SENT mailbox

[quantranhong1999]

Why

A task to build auto-complete index from the addresses that users sent mails to in the SENT mailbox.

Would help us in several cases:

• We can not reindex the contact indices.
  e.g. [Bug] No autosuggestions for recipient address tmail-flutter#3229. Rational: TWP stg/prod shared the same contact indices with tmail.linagora.com -> we needed to split the contact indices -> no data in the new TWP contacts indices yet, while we can not just reindex them using the tmail.linagora.com data...

• We migrate users' mailboxes from another system.
  => It would be helpful IMO to build auto-complete data from users' existing SENT emails which would help users have a better uninterrupted experience

How

Write a TMail task that:

• traverse users' SENT mailbox
• for each MessageResult get Headers, then parse the contacts from TO, CC, and BCC headers using Mime4j e.g. MimeMessageHeadersUtil
• index the contacts using EmailAddressContactSearchEngine

DoD

Tests + documentation for the task

[quantranhong1999]

e.g. linagora/tmail-flutter#3229

Solved this issue using Reindex a subset of documents.

{
  "source": {
    "index": "user_contact_v2",
    "query": {
      "wildcard": {
        "accountId": {
          "value": "*@stg.lin-saas.com"
        }
      }
    }
  },
  "dest": {
    "index": "twp-mail-stg-user-contact"
  }
}

[chibenwa]

Love the idea, I moved this into Todo as we could leverage this on the MU project.

Cc @guimard

[chibenwa]

Second thought about it...

What we shall do is not build directly the autocomplete DB but rather ingest those contacts onto OpenPaaS (by sending AMQP messages onto the collector:email queue)

#311 will then make sure we get those onto the auto-complete search engine.

Cc @HoussemNasri

[chibenwa]

We shall also leverage something like a bloom filter in order not to report duplicates.

[Arsnael]

What we shall do is not build directly the autocomplete DB but rather ingest those contacts onto OpenPaaS (by sending AMQP messages onto the collector:email queue)

That works in the case we use tmail with openpaas... but what if we don't?

[chibenwa]

That works in the case we use tmail with openpaas... but what if we don't?

Environments with OpenPaaS: CNB, MU, linagora.com
Environments without OpenPaaS: TWP

Then we would need two tasks (or at least 2 modes for one task) in order to be able to perform either one or the other.