Benchmark: create a "billion row challenge" benchmark

[samber]

It will be much slower than most of the results in the leaderboard, because of internals, but this is a good way to demonstrate samber/ro.

https://github.com/gunnarmorling/1brc

[kosiew]

Hi @samber 👋 — just wanted to confirm if my understanding of the issue is accurate:

The current implementation of samber/ro adds noticeable runtime overhead per emitted value due to the internal lo.TryCatchWithErrorValue and concurrency abstractions, which slows down high-volume pipelines (like the 1M row benchmark).

The issue is to try to improve the benchmark.

Is this a correct summary of the situation?

[samber]

First, we need to write the code to evaluate ro in this benchmark.

Then, find any way to improve the speed 😅