Especially for personal projects where you don’t need to show Google Analytics numbers to your company and stuff, I really think that for tracking users activity, what they click and when etc. Pheonix LiveView is nice.
For example, more and more people are using adblockers, so, the Google Analytics numbers are screwed, which is a non-issue with a server side solution.
But my concern is about performance. Logging every action in the database could be a lot after time and with a lot of users. What is your experience when it comes to performance and logging every action/click/pageload of users, or at least, registered users?
Or should I use some external open source server side solution?
Google Analytics not only can be used from client-side, but also from server-side.
Check out Server-Side Tracking in Google Tag Manager. In this way, you don’t store any action in your database, but the online service, just like using client-side tracking.
If you don’t have millions of users every month, a naive approach with a regular Postgres or MariaDB instance will probably handle your logs just fine. If you don’t care about permanently persisting these logs, you could also build something using ETS tables.
If you want something more sophisticated built in Elixir, you could give Plausible a try
Seems like too much hassle. At that point I would rather opt for something completely opensource like, I don’t know… I mean why not to write your own tracking for specific things you need and dismiss everything else?
I reiterate, I am talking mostly about personal projects, not about commercial businesses where people EXPECT Google Analytics so they can feel well and talk about with their managers and marketers during lunch.
Interesting, so via ETS I could store stuff in memory and perhaps once a day or week (I have no idea how much stuff it can hold) do a dump in a file on a disk that can be accessed via admin or put offline when necessary.
It seems nice, I will explore it further and perhaps abandon the idea of storing the clicks and actions of users in the PosgreSQL database altogether. Storing analytics in database seemed like too much bloat when what you need most of the time is just real time overview of what is happening now on your page and perhaps a few hours or days back. And the long term ego-boosting looks at the page views you do perhaps once a day or a few times a week - so, it can be compiled from files on the disk when you need it.
Thank you very much for the idea of using ETS for analytics.
