Running 1BRC was immensely fun, I learned a ton from it. Had you told me before how far the community would be able to push this, I'd have not believed you.
One main take-away for me was that you could improve performance by one order of magnitude over the baseline basically by just doing a good job and avoiding basic mistakes. The resulting code is still well readable and maintainable. In most scenarios, this is where you should stop.
If you want to improve by another order of magnitude (like leaders in the challenge did), code becomes completely non-idiomatic, super-dense, and hard to maintain. You should go there only where it really, really matters, like when building a database kernel for instance. Or well, when trying to win a coding challenge ;)
Some more resources for those interested:
* Blog post with the results: https://www.morling.dev/blog/1brc-results-are-in/
* Show & Tell, featuring implementations in languages other than Java: https://github.com/gunnarmorling/1brc/discussions/categories...
* List of many more blog posts discussing 1BRC in different languages: https://github.com/gunnarmorling/1brc?tab=readme-ov-file#1br...
* 3h deep-dive into the implementation techniques by Thomas Würthinger and Roy van Rijn, two of the top participants of the challenge: https://www.youtube.com/watch?v=_w4-BqeeC0k