ripgrep is slower than grep when searching for whitespace between words guarded by unicode word boundaries #1760
Description
What version of ripgrep are you using?
$ rg --version
ripgrep 12.1.1
-SIMD -AVX (compiled)
+SIMD +AVX (runtime)
How did you install ripgrep?
# pacman -S ripgrep
What operating system are you using ripgrep on?
$ uname -mr
5.9.9-arch1-1 x86_64
Describe your bug.
When searching a file for a pattern of the following form, ripgrep is significantly slower compared to grep, unless --no-unicode is used.
\bfoo\b \bbar\b
Removing any of the \bs above or removing the space in the middle removes this problem. The issue can also be reproduced after replacing the space with other whitespace patterns such as \t and \r.
What are the steps to reproduce the behavior?
$ git clone git@github.com:git/git.git
[...]
$ cd git
$ git ls-files | xargs cat > foo
[...]
$ ls -alh foo
-rw-r--r-- 1 user user 36M Dec 11 03:06 foo
$ file foo
foo: UTF-8 Unicode text
$ time grep -ic '\bfoo\b \bbar\b' foo
114
________________________________________________________
Executed in 79.27 millis fish external
usr time 66.16 millis 1167.00 micros 65.00 millis
sys time 13.09 millis 106.00 micros 12.98 millis
$ time rg -ic '\bfoo\b \bbar\b' foo
114
________________________________________________________
Executed in 1.60 secs fish external
usr time 1589.21 millis 1117.00 micros 1588.09 millis
sys time 6.66 millis 0.00 micros 6.66 millis
With --no-unicode, ripgrep is as fast as grep or faster.