The theme language affect search result?

[halx99]

Hello, I have question about search zh with en.

search words: set_keepalive

There are to revision docs urls:

1. https://docs.yasio.org/zh_CN/0.0.1/
   Can't search chinese mix with english

2. https://docs.yasio.org/zh_CN/latest/
   Can search chinese mix with english

the different with above two doc sites config is theme language: 0.0.1 is 'zh', latest is 'en'.

[squidfunk]

Search works for the second link you posted!? I can search in English and Chinese.

[halx99]

Yes, because I set theme lang to 'en' for second link: https://github.com/yasio/docs-936/blob/master/mkdocs.yml#L9

And the first link theme lang is 'zh': https://github.com/yasio/docs-936/blob/0.0.1/mkdocs.yml#L9

[squidfunk]

Could you try switching ja and en in the search plugin configuration and see if that resolves the issue:

plugins:
  - search:
      lang:
        - ja
        - en

[halx99]

My mkdocs.yml file already have follow config:

plugins:
  - search:
      lang:
        - ja
        - en

Please see:
https://github.com/yasio/docs-936/blob/0.0.1/mkdocs.yml#L56

But still doesn't work.

[squidfunk]

1. Yes, it's not perfect and probably never will be. Material for MkDocs supports more than 40 languages, it's very hard to make sure everything works perfectly.
2. Please see configuring the search tokenizer - it may solve both issues.

[halx99]

Thanks, I configure the search tokenizer, and the highlight issue solved, but words still break with '_':

plugins:
  - search:
      separator: '[\s\-\.]+'

[halx99]

When I change the file

mkdocs-material/src/partials/languages/zh.html

Line 34 in 5c36dfb

"search.config.pipeline": "trimmer, stemmer",

search.config.pipeline as same with en.html, then the mix search work properly.
So could we provide a configuration value search.pipeline for user, just like search.separator?

[squidfunk]

You can add the underscore to the regex:

plugins:
  - search:
      separator: '[\s\-\._]+'

[halx99]

Actually, I don't want word stop with character _, it's useful for API documentation, such as c++ opensource library may contains many APIs with character _

[squidfunk]

Chinese search support is still experimental, as we're using the Japanese language stemmer for Chinese. lunr-languages doesn't offer a dedicated stemmer for Chinese, but Japanese works reasonably well. The following configuration was reported to work for both Chinese and English:

theme:
  language: zh
plugins:
  - search