thank you for your detailed article on this subject and one suggestion you could have given some details and examples on references in migration files,
i did whatever you guided on the article everything in 30 minutes and i spent a week to fix references part in migration file, i was missing to add “type: :uuid”
add(:user_id, references(:users, type: :uuid, on_delete: :delete_all), null: false)
but finally i fixed it
ambareesha7
wmnnd
egze
If you are installing Phoenix JS libraries from NPM, you might find this post useful 
stefanluptak
Thanks 
Sebb
We learn best from errors,but obviously most blog-posts are shiny success-stories. So thanks for your courage to share this!
derek-zhou
So dozens of updates per second through pubsub is ok, it is just the live view process cannot keep up, right?
Would it make sense to store the participants info in ETS and just send a simple :updated message to the liveview, and within the liveview it can selectively peek forward the message queue and collapse all the :updated messages, grab the participants from ETS then render just once?
wmnnd
The issue really ended up being a combination of these things: payloads too large, updates too frequent, processing in the LV process too slow.
And while I haven’t gone for ETS just yet, I did change it from {:participants_updated, participants_list} to just :participants_updated plus an explicit fetch from the LV.
derek-zhou
Once your message is converted to just a notification, you can do a receive loop and cull all the participants_updated messages in the queue, then fetch the list once, render the html once. If you have the actual page render and the diff sent, the bottleneck should not be the client side anymore. The inbox of the LV process should be under control due to the smaller messages and the message coalescing.
wmnnd
I found throttling the update messages at the sender level to be a nicer solution. Instead of having the LV account for an excessive amount of messages, I prefer making the sender “play nice”
dorgan
I wrote about the challenges I faced when trying to programmatically modify the source code and how that led to the creation of Sourceror