Integrated search engine

[amirouche]

I think it will be better to have a search engine integrated to the static version of the website. The problem is that the index is too big to be built by the user even for superuser (20MB). lunrjs provide numerous entry points. Here is the solutions that remains to be explored:

• build the index using lunr and store in the build directory. Load the index from the build directory.
• If the above doesn't work because the index is too big (and it might be too big for slow machines). Create a new index for lunr.js that fetch store items lazily as required by the query engine this will save a lot of memory.

I think it's better to contribute back to lunr.js which is an established project regarding this topic. I had a look a look at other solutions but they more complicated and we only require english support.

Maybe it's better to thing I18N and go with a FTS solution that can handle mutliple languages. Again, I think it's better to have search integrated to the static part of the ZIM file that have search engine built inside the kiwix reader which breaks UX.

[kelson42]

Did you have had a look to the solution used for Gutenberg? I would recommend to get the static pages done and correctly linked. Make a first version and then work on this. Meanwhile the reader fulltext search engine should do the job.

[amirouche]

This is kind of complex. I'd like to do a first release before working on that.

[dattaz]

As discuss with @kelson42 search in zimlib, will, for the moment and for the use case, be enough