fix for #1588

[tcoopman]

This is a proposal fix for #1588. Hyphens now are found by the index and get a higher score.

These 2 lunr issues are relevant to understand the fix:

• Searching terms contain hyphen vs no hyphen olivernn/lunr.js#296
• Query term separator not in sync with tokenizer separator  olivernn/lunr.js#493

Some remarks about the solution:

• I've exported a __createIndex functions to be able to test it (createIndex uses globals)
• I've used the searchNodes from the live view docs, which is overkill and can be changed, but before I did, I wanted to make sure that this is the direction you want.
• I've got no idea on how to build/setup a version of ex_doc and test on the real docs to see how it actually works. If that's something I can easily do, let me know and I'll try it. (so I don't know if I broke anything...)
• I tried to maintain the coding style of the javascript files, but as there is no formatter config, I'm not sure it's perfect.

[josevalim]

Hi @tcoopman, awesome job!

Given this is mostly configuration of the search engine, I would say we can go ahead and skip the tests and only focus on the fix. Can you please amend accordingly? Thank you!

[tcoopman]

Like this?

[josevalim]

@tcoopman can you also please follow the style of this PR?

14f0efc

Note that we need to register the function also when loading a serialized index, otherwise the serialized index does not work. :)

[tcoopman]

I think it should be good.
One remark the hypenSearch adds the option to search for "phxupdate" now as well. So it should have better results for "phx-update" and "phxupdate".

[josevalim]

💚 💙 💜 💛 ❤️