Re: [Tracker] Wish list: ignore certain types of files in certain directories

[Sam Thursfield <ssssam gmail com>]

Hi Bert,

On 9/19/17, Bert Bos <bert w3 org> wrote:
> Here is my use case:
>
> Tracker indexes my photos & other images and my music collection. In the
> directories containing my music there are often image files with cover art.
> Most of them are called "cover.jpg”. I don’t want those to show up in my
> photo collection, because there are too many and they make it difficult to
> find my own images.

This makes a lot of sense, I have the same issue.

I would suggest that the way to solve this is to have another step of
processing, before metadata extraction. Currently we do this:

1. tracker-miner-fs: read basic info for each file (filename, mtime,
MIME type, etc.)
2. tracker-extract: read content-specific metadata for each file

That's all we do, and so we make a bunch of assumptions that don't
actually hold such as "every picture file is a photo", "every audio
file is a piece of music" and "every text file is a document that we
should index."

If we had an intermediate step, we could look at each directory once
all the files in there have been crawled and make a guess about
whether the directory has a special 'role', such as being a music
album, or being a source code repo.

I wouldn't want this to get too complicated as it's always going to be
a bit unreliable, but there is a simple rule we could add to solve
this case:

 * if a directory contains one or more music files and some image
files matching 'cover*.jpg'; that directory is a music album and those
image file are album art, not photos

Sam