Hayden is a reimplementation of Asagi in .NET Core for ultra-low resource usage, both memory and performance.

• Much smaller memory consumption than Asagi.
  • For comparison, Hayden requires 10MB of working memory to archive a single board (including all archived threads), while Asagi consumes several gigabytes to do the same. (Multiple boards seem to also only require 10MB, subjec to further benchmarking)
• Efficient usage of API calls, allowing the scraper to be API compliant while also not missing any posts.
  • It uses the threads.json and archives.json endpoints to determine which threads have been updates, instead of continously polling.
  • Once the thread cache has warmed up, it will only update threads that have been detected as changed. On /g/ for example, this amounts to about 4 threads updated per 30 seconds. There's a lot of room to cram other boards in there, as a result.

To be fleshed out. Currently it's executed as hayden <board to archive> <directory on disk to store data>

I needed a tool of my own to download threads from 4chan, since the only other ones that exist either don't work or are written in python. I found out about the wasteful resource usage of Asagi and (currently in the process of creating) a backend to support Asagi to help out 4chan archive owners who are currently crumbling under having to require several tens of gigabytes to keep Asagi archiving every board.

GC? Yes. Wasteful memory usage? Obviously not.

GC slowdowns are avoided due to me keeping the object limit down by only keeping in memory data that I absolutely need to cache, and strategically declaring some data strucures as structs. This also contributes to the very low memory usage.

Not to mention C# supports asynchronous code, where-as Java does not. I believe this is where the majority of wasted memory is in Asagi, as it needs to spin up a new process thread for every imageboard thread.

I was listening to the Doom 2016 soundtrack as I was programming this.