Re: File search

[raingloom]

On Fri, 21 Jan 2022 10:03:43 +0100
Ludovic Courtès <ludo@gnu.org> wrote:

> Hello Guix!
> 
> Lately I found myself going several times to
> <https://packages.debian.org> to look for packages providing a given
> file and I thought it’s time to do something about it.
> 
> The script below creates an SQLite database for the current set of
> packages, but only for those already in the store:
> 
>   guix repl file-database.scm populate
> 
> That creates /tmp/db; it took about 25mn on berlin, for 18K packages.
> Then you can run, say:
> 
>   guix repl file-database.scm search boot-9.scm
> 
> to find which packages provide a file named ‘boot-9.scm’.  That part
> is instantaneous.
> 
> The database for 18K packages is quite big:
> 
> --8<---------------cut here---------------start------------->8---
> $ du -h /tmp/db*
> 389M    /tmp/db
> 82M     /tmp/db.gz
> 61M     /tmp/db.zst
> --8<---------------cut here---------------end--------------->8---
> 
> How do we expose that information?  There are several criteria I can
> think of: accuracy, freshness, privacy, responsiveness, off-line
> operation.
> 
> I think accuracy (making sure you get results that correspond
> precisely to, say, your current channel revisions and your current
> system) is not a high priority: some result is better than no result.
>  Likewise for freshness: results for an older version of a given
> package may still be valid now.
> 
> In terms of privacy, I think it’s better if we can avoid making one
> request per file searched for.  Off-line operation would be sweet, and
> it comes with responsiveness; fast off-line search is necessary for
> things like ‘command-not-found’ (where the shell tells you what
> package to install when a command is not found).
> 
> Based on that, it is tempting to just distribute a full database from
> ci.guix, say, that the client command would regularly fetch.  The
> downside is that that’s quite a lot of data to download; if you use
> the file search command infrequently, you might find yourself
> spending more time downloading the database than actually searching
> it.
> 
> We could have a hybrid solution: distribute a database that contains
> only files in /bin and /sbin (it should be much smaller), and for
> everything else, resort to a web service (the Data Service could be
> extended to include file lists).  That way, we’d have fast
> privacy-respecting search for command names, and on-line search for
> everything else.
> 
> Thoughts?
> 
> Ludo’.
> 

One use case that I hope can be addressed is TeXlive packages. Trying
to figure out which package corresponded to which missing file was a
nightmare the last I had to use LaTeX.