gregblake

Connect::Message.with_context
returns a message, and a number of messages that preceed and succeed it, in the same conversation
  public rooms
    behaves like a conversation that the user has access to
      when the preceding & succeeding message count are less than the number of messages in the convo
        returns the selected msg, along with the correct # of messages, to provide context for the messsage
      when the combined preceding & succeeding message counts are greater than the # of msgs in the convo
        returns the selected msg, along with every other message in the conversation
      when the preceding_message_count is zero
        returns 0 preceeding msgs, the selected msg, & the correct number of succeeding messages in the convo

gregblake

ElasticSearch API Endpoint

This endpoint will allow clients to search messages within a given conversation.

1. Request params (room/group): :to_id, :to_type, :query.
2. Request params (user): :conversation, :query.
3. If the query parameter is omitted from the request, it should behave exactly like the messages endpoint. Similarly, if the toggle and/or permission are disabled, it'll behave just like the messages endpoint does, without search filtering.
4. If the necessary parameters are included: return a 200, with a collection of messages that match the query. The response JSON should match the existing format exactly.

Other considerations

gregblake

Steps to Upgrading

1. Upgraded cluster
2. Upgrade gems from 2.0.
3. Audit application for breaking changes, and update them accordingly.
4. Refresh indices (?)

Other Considerations

• It looks like we can only upgrade to v6, and not v7 yet. See compatibility chart for the elasticsearch-rails gem.
• Is the messages table being indexed in production now? If so, will this be compatible after the upgrade? Will we need to refresh the index?

gregblake

What did I do yesterday to help the team meet the sprint goal?

• Completed the story that adds a new composite index to the connect_rooms table. This is in production. No new errors in Sentry.
• Began working on two new stories:
  1. Add two new feature toggles for Connect search.
  2. Add a new permission for Connect search.
  • Both of these PR's received code review and testing approval. I merged both PR's in the afternoon. Another team reverted a PR shortly after that, so they shipped via emergency deployment. But I followed that deployment into production. No new errors in Sentry.
• Worked on the story to optimize the connect_room_users query. A PR is open, and the feature is nearly completed.
• Code review.
• Chatted with Jenny about upgrading the mobile app from RN v0.59 to v0.60/v0.61.
• Attended the All Hands meeting.

gregblake

❯ echo "[RBLS-39]" | sed 's/^\[\(.*\)\]$/\1/'
RBLS-39
❯ echo "[INC-3399993]" | sed 's/^\[\(.*\)\]$/\1/'
INC-3399993

gregblake

Uncaught TypeError: Cannot convert undefined or null to object
    at Function.keys (<anonymous>)
    at Message.renderReactions (Message.jsx?dd7c:405)
    at Message.render (Message.jsx?dd7c:445)
    at eval (ReactCompositeComponent.js?cd59:793)
    at measureLifeCyclePerf (ReactCompositeComponent.js?cd59:74)
    at ReactCompositeComponentWrapper._renderValidatedComponentWithoutOwnerOrContext (ReactCompositeComponent.js?cd59:792)
    at ReactCompositeComponentWrapper._renderValidatedComponent (ReactCompositeComponent.js?cd59:819)
    at ReactCompositeComponentWrapper.performInitialMount (ReactCompositeComponent.js?cd59:361)
    at ReactCompositeComponentWrapper.mountComponent (ReactCompositeComponent.js?cd59:257)

gregblake

Android

App State	Behavior after Tapping Notification
Foreground	Same as prod
Background	Route to conversation
Not Running	Route to conversation

Some/all Android devices don't receive push notifications when the app is in the foreground, so that's why I mentioned "same as prod" for Android apps when the app is in the foreground).

gregblake

This query optimization can be recreated locally, by running the following and grepping the logs (comparing the master branch to this feature branch).

Prior to this PR, the number of queries issued was dependent upon the number of rooms in the collection (N+1). In production, the collection often contains many rooms. Now, only two queries are issued, no matter how many rooms are in the collection.

From the Rails console:

rooms = Connect::Room.where(private: true, archived: false).first(10);
rooms.each do |room|
  Connect::Message.create(from_id: 618, to: room, body: "u#all");
end

gregblake

ActiveRecord::Base.establish_connection
ActiveRecord::Base.connection.execute("SET FOREIGN_KEY_CHECKS=0;")

ActiveRecord::Base.connection.tables.each do |table|
  next if table == 'schema_migrations'

  case ActiveRecord::Base.connection.adapter_name.downcase.to_sym
    when :mysql2 || :postgresql
      ActiveRecord::Base.connection.execute("TRUNCATE #{table}")
    when :sqlite

gregblake

What did I do Wednesday to help the team meet the sprint goal?

• Finished the “ Image partially grays when loading” story.
  • This is in production.
  • After our standup conversation, Stefan added a follow up story to address the remaining edge case on the rn-fetch-blob bug: https://www.pivotaltracker.com/story/show/169042189
• Picked up the “wrong label & color when removing a bookmark” story from the backlog.
  • This is in production.
• Picked up the “unable to view a user's contact details” story from the backlog.
  • This one has code review approval, and is in testing.
• I did some code review.