Seems I can't reply to your other comment, so I'll reply here. How can you say that my naive implementation is not naive? What is not naive about it? It's very naive. It's basically the same naive code that you were writing, but in actual idiomatic Rust with the linting issues fixed.
Using a `lines()` approach is naive because that allocates owned heap-allocated strings. An optimal, non-naive solution would not use heap allocation for buffering lines but use a stack-allocated array. That alone would bring significant speedup versus the line approach.
As for ripgrep, it's a pretty comprehensive application that makes use of SIMD/AVX so it's only natural that it's fast.
Your solution is not naive in my opinion because you set the size of the buffer and use map/filter but ok... Let's check. Your solution is the slowest from all the solutions.
It took 4.6s which only confirms what I wrote on beginning when we started this discussion. Perf counters for your solution in here:
Using a `lines()` approach is naive because that allocates owned heap-allocated strings. An optimal, non-naive solution would not use heap allocation for buffering lines but use a stack-allocated array. That alone would bring significant speedup versus the line approach.
As for ripgrep, it's a pretty comprehensive application that makes use of SIMD/AVX so it's only natural that it's fast.