Hacker News Reader

Comment by chaps | Hacker News Reader

Nice! Similarity searching is the bane of my existence, but also gives life to so many of the things I work on. 200x sounds wonderfully fast.

Does this do levenshtein distance? Didn't see it in the main document.

Edit: also, THANK YOU for adding a parameter to specify # of threads. Wish this was a wider standard practice.

View on HN

The author is also responsible for the StringZilla library that includes levenstein distance: https://github.com/ashvardanian/StringZilla

He-hey! For Levenshtein distance check out the development branch of StringZilla. Hopefully will release this week :)

Oh, and for multithreading, USearch might be more practical.

Heck yeah, thanks.

  If the first 4 bytes of the string  are the same, the strings are likely to be equal. Similarly, the first 4 bytes of the strings can be used to determine their relative order most of the time.

That's a fun heuristic.

Yes, it was very practical in most cases, but v3 only uses it for short needles with SWAR. SIMD variants will use Raita-style heuristics.

OMP_NUM_THREADS=n is pretty standard in HPC workloads

Huh, thanks. I'll check it out and see if I can drop that in.

Reply on news.ycombinator.com