Ask HN: How to (instantly) search by filename in Linux?

9 points

11 years ago

I'm hoping someone has some advice or a tool to recommend for this particular problem. I have many tens of thousands of (binary) files in a few directories spread over a few filesystems. I'd like a way to search by filename (either pattern or regular expression).

Right now I am using a quick script to compose a somewhat complex `find` command. This is otherwise effective but quite slow. Each query takes anywhere from 1-20 seconds.

An alternative that I pursued was `mlocate`. A daily cron (or systemd timer as the case may be) script generates a database for each directory (e.g. `updatedb --database-root DIRECTORY_A --output DATABASE_A.db --require-visibility 0`). Then to search `locate --basename --database DATABASE_A.db:DATABASE_B.db:DATABASE_C.db PATTERN`. However unlike `find`, `mlocate` does not offer an ignore or '!'. I suppose I could then strip ignored paths from the output with another tool, but things are starting to get pretty hackish at this point.

Perhaps there is something like Bitbucket's Quick File Search [1]?

Or something along the lines of etsy's hound [2] but for files rather than code of course. I've been using hound for instant search of all my repos and it is quite incredible.

[1]: http://blog.bitbucket.org/2013/02/07/introducing-quick-file-search/

[2]: https://github.com/etsy/hound

15 comments