tagfarm

Version 0.4 | Entry @ catseye.tc | See also: shelf ∘ ellsync ∘ yastasoti

---

tagfarm is an ultra-lightweight filesystem-based categorization system for arbitrary files.

Motivation

Most filesystems that we use in the modern day are hierarchical filesystems, where directories contain files and nested directories. A significant limitation of these filesystems is that any given file can only reside in one directory. But most files are best described by classifying them in more than one category. So directories don't serve very well as categories -- if you have a JPEG of the Mona Lisa, should you put it in Portraits or Works by Leonardo da Vinci or Enigmatic Smiles?

A solution used on many blogs and wikis is to allow articles to be labelled with a set of "tags", and to support listing all articles with a given tag in "tag index".

tagfarm implements the same idea on a local filesystem. Each tag is implemented as a directory containing symbolic links to the files that have that tag. And this is all it consists of.

There are several advantages to this. There's no metadata to go out of sync, no database engine to install and maintain. When you move files around, you can just run tagfarm repair to rewrite the tag links. You can treat the tag links as you would any other file — for example, you can remove a tag from a file just by deleting the tag link.

There are also some disadvantages, of course. Primarily, any limitations that your filesystem has are also going to be imposed on the categorization system. So, for example, if your OS has performance problems listing 10,000 files in a single directory, it would also have problems listing 10,000 files tagged with Enigmatic Smiles.

Quick start

Make sure you have Python (2.7 or 3.x) installed, clone this repository, and put its bin directory on your executable search path. You can then run tagfarm from your terminal.

New in 0.4 - Nautilus Extension

You can also use tagfarm from Nautilus (the default file browser of the GNOME Desktop, which is the default desktop environment of Ubuntu). After putting bin on your search path, run these commands from the root directory of this repository to install the necessary and restart Nautilus:

sudo apt-get install python3-nautilus
mkdir -p ~/.local/share/nautilus-python/extensions/
cp script/tagfarm_nautilus.py ~/.local/share/nautilus-python/extensions/
nautilus -q

After installing, the context menu (brought up by right-clicking on a file) will contain new menu items to add a tag, remove a tag, and repair tag links.

Overview

Media tree

tagfarm operates on a part of your filesystem it calls the media tree. There may be multiple media trees in your filesystem. The topmost directory of a media tree is called the media root and it is identified by the fact that it contains a directory called by-tag. When tagfarm is started, it finds the media root it will operate on, by looking for the by-tag directory, first in the current directory, then in every successive parent directory thereof. If it reaches / without having found a by-tag directory, it exits immediately with an error code.

One constraint that applies to the media tree is that every file in it should have a unique name. This allows tagfarm repair to recreate fix broken tag links when a file is moved around inside the media tree.

tag and untag

You can add a tag to one or more files with the command

tagfarm tag <tagname> <filename> [<filename2>...]

You can then list the contents of by-tag/<tagname> and see there are symlinks there.

You can remove a tag to one or more files with the command

tagfarm untag <tagname> <filename> [<filename2>...]

showtags

You can list the tags that have been applied to one or more files with

tagfarm showtags <filename> [<filename2>...]

There are several options that can be supplied to influence the format of the listed tags. See tagfarm showtags --help for these.

repair

If the tagged files are moved around inside the media tree, the symbolic links that represent their tags will break. Assuming that the files, after having been moved, keep their filenames unchanged, tagfarm can repair the broken links with the command

tagfarm repair

By default, tagfarm will only attempt to repair broken links if they are actually symlinks (and not, for example, regular files) and are actually broken. To have it replace all files it happens to find in the tag link directory, pass --force-relink. This is occasionally handy for converting a directory full of copies of elsewhere- existing media files into links. In conjunction with this, --restrict-to-tag may be used to name a single tag, and this operation will be applied only to that tag.

tagfarm repair will also replace any links it finds that have absolute target paths, with ones with relative target paths, even when the link is not broken.

tagfarm repair will also, when processing a link whose name is like Link to xyz, rename it to simply xyz. This is to handle the case where links are manually created in a tag directory using a tool such as Nautilus (the default file browser of the GNOME Desktop).

rename

tagfarm rename <oldfilename> <newfilename>

Renames the file, like mv, and also updates any tags that might be on it.

collect

tagfarm collect <tagname> <destdir>

Convenience command to move all the files with a given tag into a destination directory. If the destination directory does not yet exist, will be created. The tags of the files are not changed. At the end, tagfarm repair is called to update the tag links.

Other operations

The advantage of tagfarm being ultra-lightweight is that if there is something that it does not directly support, it's often easy to accomplish it by simply issuing some conventional commands to alter the filesystem. For this reason, some functionalities you might expect to exist, don't have specific tagfarm commands implemented for them.

For example, to rename a tag, one needs only to rename the directory that contains the tag links. For example:

mv by-tag/airplane by-tag/aeroplane

TODO

Better handling of cases where the target being linked is itself a link.

Set-theoretic queries on tags (e.g. tag all files with X or Y and not Z with a new tag T).