Closed rudokemper closed 1 month ago
This comment is just FYI and I don't think you should go divert from the existing plan. But I read https://taylor.town/pg-task and really liked it -- it roughly describes a way that we could drop Azure Storage Queue entirely. And in fact I've followed his pattern in something on the acoustic side, works pretty well!
Just skimmed it and it absolutely resonates with me - and I think I was hoping for something like this in the initial iteration of the design doc.
Adopting this approach will make it much easier to make the map-packer + mapgl-tile-renderer toolchain run anywhere.
But yes, this far into work towards an MVP, we probably shouldn't divert from the plan (but judging by the implementation code, I also don't think it would be a ton of work to drop ASQ for this approach, especially if you have experience on the acoustics side!)
One reason I suggest holding off is it will require refactoring / splitting your SQL tables. This wasn't obvious to me at first but I quickly found out when implementing. Basically the task queue table needs to be only a task queue, and cannot power anything else, e.g. the list of maps in the webapp.
n/t