Removing diacritics when tokenizing for search does not work retroactively?

AI-generated summary

The discussion revolves around the removal of diacritics when tokenizing for search in Discourse, a feature recently implemented by zogstrip. However, danekhollas reported that the feature does not work retroactively after upgrading via the upgrade manager UI.

sam suggested that bumping the search version would force a reindex, which would take a few days to complete. Alternatively, zogstrip and sam recommended using a rake task (rake search:reindex) to manually rebuild the index.

danekhollas successfully applied this solution, confirming it works as expected.

Continuing the discussion from Discourse should ignore if a character is accented when doing a search:

@zogstrip implemented a change to search which should now ignore accents. I did an upgrade and tried it out but it does not seem to work. Is it possible that it will only work for new posts? If so, is it possible to apply the fix retroactively i.e. rebuild the PG index?

Or maybe I am doing something wrong? btw: I upgraded via upgrade manager UI.

Thanks!

We need to bump the search version to force a reindex, I guess it is overdue since we had multiple search improvements lately

Once we bump the version it will take a few days for index to rebuild

Isn’t there also a rake task if they want to do it manually in the meantime?

Yeah there is a search reindex task

If you don’t want to wait for @sam’s changes, SSH into your server and then

cd /var/discourse
./launcher enter app
rake search:reindex

Wohoo, works like a charm, thanks so much! :heart: