How to express the -w option in a regular expression string？

[XhstormR]

I have a sample like this, the word 重庆 is chinese:

   重庆  a
重庆
,重庆,

a重庆a
重重庆庆

I can use -w option to find word 重庆 like this rg -w "(?-u:\xe9\x87\x8d\xe5\xba\x86)" test.txt :

1:   重庆  a
2:重庆
3:,重庆,

But somehow I can't use options, I can only use regular expression strings, so i use command rg "(?-u:\b\xe9\x87\x8d\xe5\xba\x86\b)" test.txt, but the output is incorrect:

5:a重庆a

How to express the -w option in a regular expression string？

[BurntSushi]

Good question!

Why are you using the raw UTF-8 encoding like that? When you use \b when Unicode mode is disabled, then it's an ASCII word boundary, not a Unicode word boundary. So it will not work correctly with text that is not ASCII.

Firstly, with the -w flag, you can invoke ripgrep with your Chinese characters directly:

$ rg -w '重庆' haystack
1:   重庆  a
2:重庆
3:,重庆,

Similarly, without the -w flag:

$ rg '\b重庆\b' haystack
1:   重庆  a
2:重庆
3:,重庆,

If you want to continue using the raw UTF-8 encoding for some reason, then you may do so, but only disable Unicode around the portion of the regex that is matching raw bytes:

$ rg "\b(?-u:\xe9\x87\x8d\xe5\xba\x86)\b" haystack
1:   重庆  a
2:重庆
3:,重庆,

And finally, for completeness, note that the -w/--word-regexp flag is actually not implemented with \b(regex)\b. It is implemented as (?:\A|\W)(regex)(?:\z|\W). This has slightly more intuitive results in some cases. See #389 for more details and examples. But, if you need to enter the regex directly, then you should probably just use the \b(regex)\b variant since the more complex variant will potentially highlight the non-word characters on either side of the match.

[XhstormR]

Thanks, this is very helpful.