Handling of shared folders in fulltext search

🍱 Features & apps fulltextsearch
1

I have a quite large folder (about 18000 files) on my Nextcloud server which is shared with about 24 people. However the folder contains files which are not relevant for fulltext search but more a shared library so everyone has access to it. I added a file named .noindex to the root of the folder to exclude it from indexing and search.

Well - this completely ignored :frowning: The fulltext search happily starts indexing 18000 files for every single user with whom the folder is shared. So in the end the index reads 432.000(!) files and the process takes multiple days(!) to complete (with about 1-2 files per second I estimate about 3-5 days until the indexer is finished - so far it only finished about 3 users after about 5 hours. So I don’t expect it to finish very soon…)

Besides the question why .noindex does not work I also wonder, why the fulltext search does not recognized the fact, that file comes from a shared folder and only needs to be indexed once. Even when taking into account that the same shared folder might have a different local path for the users you may still only need the index once with a user specific reference where to find that file when that specific user searches for it.

Did I miss anything? Is this the intended behaviour?

1 Like
2

Hello. Did you uncover a solution to this problem? I observe similar, however I do want my common files to be indexed so everyone can search. I’d just like them indexed only once.

3

Hmm. Sounds like this open bug report. Perhaps give that issue in GitHub a thumbs up (so it is clearer it’s not an isolated report). From what I see, the code supporting .noindex is still in place.

4

Thanks. Not sure my issue is the same as that one as I do want my common/shared files indexed, however the ability to exclude subtrees from the index is definitely a commonly useful thing to support.