This is a fuse-based virtual file system which allows use to attach tags to files and then navigate the tag cloud just like usual directories.
Two principal limitations: you cannot tag directories, and all the files you wish to access through one mount point must reside in one "source" directory somewhere else. Or at least symlinks to them, which is still cumbersome. This is ok if you are going to access the collection of papers or books, which was my primary purpose for trivialfs: you just never need to go to this directory.
Assume you mounted trivialfs to ~/tags.
ls ~/tags/algebra/books
will show you the list of all files which are tagged algebra and books. Similarly, ~/tags/algebra/books/finished gives you all files with three tags, algebra, books, finished. The directory ~/tags/books/finished/algebra is absolutely the same. Do not use find on trivialfs directories! It will do an exponential amount of work, treating all combinations of tags (which actually correspond to at least one file) as separate directories.
Subdirectories in trivialfs mounts are, as you may guess, more tags which could be added to the list of tags (like algebra and books in the example) with at least one file having all the tags.
It is impossible in current version to filter out files with tag a and not tag b beyond some awk/sed manipulations with lists of files: only intersections are supported.
Files without tags could not be accessed through the mount point in any way. Also everything in trivialfs is read-only.
You don't need to unmount and mount trivialfs manually if you added new tagged files or retagged already existing ones. Execute
touch ~/tags/reload
to apply the changes on the fly. The success is indicated by returning "File not exists" error, which is not something you are going to get from touch (sorry for this). In more details, trying to create a file named "reload" in the directory you mounted trivialfs to (not inside some tags, so ~/tags/algebra/reload will not work) leads to reparsing .tags file.
First of all, you need a separate directory with all the files you are going to access through one mountpoint. Let's assume it's ~/info-storage. The information about tags is contained will be contained in the file .tags there. It has a human-readable format, so you may just modify it as you with. It's usually more convenient to use the script trivialtags bundled together with trivialfs.
To see which tags does the file "Vakil AG.pdf" have currently, run
trivialtags "Vakil AG.pdf"
To remove all tags from this file, execute
trivialtags "Vakil AG.pdf" -
To remove just one tag, say, "ag", execute
trivialtags "Vakil AG.pdf" -ag
There is no way to remove several but not all tags from a file beyond repeating this command with different tags. It seems reasonable to assume that if you have 10+ tags on every file, you already have some way to automate tagging.
To add a tag "finished" to this file, run
trivialtags "Vakil AG.pdf" finished
To add several tags give a list separated by commas (as one of arguments):
trivialtags "Vakil AG.pdf" "book to be used again and again, nice, fundamental"
(spaces around commas will be ignored, but spaces inside the tag are nice)
If you run trivialtags without arguments, it will give you similar help text.
If the directory with all the files and .tags is ~/source and you wish to access them through ~/tags, execute
trivialfs ~/source ~/tags &
(if you need to mount only if it's not mounted yet, use something like
(mount | grep -q $HOME/tags || trivialfs /mnt/files/info-storage $HOME/tags) &
but you probably know this better than me)
Remembed to use &: trivialfs doesn't switch to a daemon state.
Run make in the directory with source files. If you are using Arch Linux, you may also want to run make dist after that to create trivialfs.tar.gz with all the binaries, modify md5 in PKGBUILD and then run makepkg to obtain pacman-installable package.