Lunr search index merging words across newlines

[bmcgough]

The lunr index in minimal-mistakes is fed using an excerpt from each page in /assets/js/lunr/lunr-store.js. The excerpt is passed through strip_newlines which removes any newlines in the excerpt, but does not replace those newlines with anything, resulting in merged words across newlines (ex: foo\nbar becomes foobar, but should become foo bar).

So our search index has many merged words.

I'm not sure where this should be fixed. One option would be to do something stupid in minimal-mistakes like newline_to_br followed by replace: '<br>', ' ' rather than the strip_newlines, but probably the "right" solution is to modify strip_newlines to replace with spaces, or add an option or new function in Liquid to do that. It looks like Liquid's replace cannot easily to used on newlines directly (which is presumably why the newline specific function exist).

What do you think?

[mmistakes]

Probably not going to modify Liquid or introduce a new function there that would make its way to Jekyll anytime soon 😉

But I think your proposal is probably the best option give what's available to us. Care to submit a pull request removing strip_newlines and using carefully constructed replaces to get around the merged words?