Question: Difference between ripgrep and fastmod? about amber HOT 4 CLOSED

Thank you for your suggestion.
I updated benchmark results with hyperfine.

rg/fastmod is fast in the case of many files,
and amber has advantage in the case of large file search.

from amber.

Nice, cool to see the benchmark. I am surprised that amber seems faster here. Maybe @BurntSushi might be interested in this.

from amber.

Benchmarks should include checks for correctness, otherwise the whole endeavor is moot:

$ time rg '検索結果' jawiki-latest-pages-articles.xml | wc -l
6455

real    1.499
user    1.060
sys     0.437
maxmem  10914 MB

$ time LC_ALL=en_US.UTF-8 grep '検索結果' jawiki-latest-pages-articles.xml | wc -l
6455

real    3.535
user    2.199
sys     1.333
maxmem  4 MB

$ time ~/clones/amber/target/release/ambs '検索結果' jawiki-latest-pages-articles.xml | wc -l
7048

real    0.796
user    6.399
sys     1.499
maxmem  10871 MB

Assuming it's a simple logical bug that doesn't impact performance, then yeah, this is a very nice validation of the idea to search a single file in parallel. If you use regexes correctly, then it might go even faster, since a regex's literal search should be better than the ones you've hand-rolled. But right now, the code is unnecessarily doing UTF-8 validation, which kills the entire enterprise. Instead, you should be using regex::bytes::Regex.

Nice work @dalance!

from amber.

@BurntSushi
Thank you for your comment.
Currently amber show line by match. So the line matched repeatedly is duplicated.
I'll fix it, and try regex::bytes::Regex.

from amber.