Tag files and folders anywhere on your filesystem, then archive them all at once, or unpack one you already have.

You walk around, press space on anything you want, press p, answer a few questions, and it builds the archive. The progress view can be sent to the background while you keep browsing. Press u on an archive instead and it unpacks, wherever you want it.

Needs Ruby 3.0 or newer and nothing else.

gem install rakpak

Then rakpak from any folder. If your shell cannot find it afterwards, the folder RubyGems puts executables in is not on your PATH; gem env shows it under EXECUTABLE DIRECTORY.

Without RubyGems, clone this repo and run ./install.sh. That copies the program to ~/.local/share/rakpak and puts a launcher in ~/.local/bin, adding that folder to your shell startup file if it is not on your PATH yet. ./install.sh --link runs straight from the checkout so edits take effect at once, --prefix /usr/local installs system-wide, and --uninstall removes it.

rakpak                    browse from your home folder
rakpak ~/Documents        browse from there instead
rakpak .                  browse from the current folder
rakpak -p ~/Documents     pack that folder: opens on the archive prompts
rakpak -p a.txt b/ c      pack several things at once
rakpak -d site.tar.gz     unpack it here and now, into site/
rakpak -d a.tgz b.zip     unpack several, each into its own folder

With no arguments rakpak opens in your home folder: its contents in the middle pane, the folder above it on the left, a preview of whatever is under the cursor on the right. Give it a folder and it opens there instead.

-p (or --pack) skips the browsing. Every path you give is tagged, the browser opens on the folder holding the first one with the cursor resting on it, and the archive prompts appear at once, starting with the kind of archive and its compression. The archive is written next to that first path. Press esc on any prompt and you are back in the browser with the tags still set.

-d (or --depack) does not open the browser at all. Each archive is unpacked straight into the folder you are standing in, the way tar would, each one into a folder of its own name: site.tar.gz gives you site/. A lone compressed file has nothing to wrap, so notes.txt.gz lands beside you as notes.txt. One bad archive does not stop the rest, and the exit status is non-zero if any of them failed.

1. Browse and tag. Vim keys or arrows. space tags whatever is under the cursor and moves down. Tags persist as you walk, so you can pick something in ~/Documents, walk to /etc, and tag more. T reviews everything tagged. t opens a queue pane on the far left that stays up while you browse. It lists each queued item by name, where it lives, and its size, with the running total on top.
2. p to pack. Choose compressed tarball (the default), plain tarball, or zip archive.
3. Set the flags. Compression method, level, and the switches for that format. Compressed tarballs default to gzip, the one every system can read.
4. Leave things out. Type names separated by commas, like node_modules, .cache, *.log. Each one is left out wherever it turns up, along with everything inside it. Or pick Tag files and directories to walk the folders you are packing and press space on the exact spots to skip; p brings you back. Skipped when only files are packed.
5. Say where. This directory, your home directory, or one you type into the field right there. ~, $HOME and absolute paths all work, and typing anything jumps to the field.
6. Name it. The extension is added for you.
7. Confirm. You see the exact command before anything runs. Only enter starts it.
8. Watch it, or press b to drop it into the background and keep browsing.

If nothing is tagged, p archives whatever the cursor is on.

u on an archive opens three prompts: which folder to unpack into, what to call the folder it makes there, and the usual confirm screen showing the exact command.

Unlike p, this follows the cursor rather than the tag set: one archive, one destination. The folder is offered as the archive's own name with the extension taken off, so site.tar.gz suggests site/, which means nothing is ever sprayed across the folder you are standing in. Type . as the name to unpack straight into the folder you picked instead, and the folder prompt takes ~, $HOME and any absolute path, so the contents can go anywhere. A lone compressed file skips the folder entirely.

If the destination already has files in it, the confirm screen says so before anything runs. A failed extraction leaves whatever it managed to write; the files that were already there are never touched.

In the job view: b backgrounds it, x cancels, esc returns to the browser, q quits.

Every run produces exactly one file.

Compressed tarball: tar piped through gzip, zstd, xz, bzip2, lz4 or brotli, giving .tar.gz, .tar.zst and so on. Each has a level, and the tar switches (permissions, xattrs, symlink handling, VCS exclusion, reproducible ordering, and so on) are on the same screen.

Plain tarball: a bare .tar, no compression. Just the tar switches.

Zip archive: a .zip written by Info-ZIP with deflate, bzip2 or store. When exactly one file is selected, the single-file compressors are offered here too, so notes.txt becomes notes.txt.gz or notes.txt.zst without a tarball around it. For a folder those are greyed out with the reason, since gzip alone cannot take a folder.

Every one of those can be read back: .tar, .tar.gz, .tar.zst, .tar.xz, .tar.bz2, .tar.lz4, .tar.br, the .tgz, .tzst, .txz and .tbz2 short forms, .zip via unzip, and .gz, .zst, .xz, .bz2, .lz4 or .br on their own.

Everything is probed at startup: PATH for the compressors, tar --version to tell GNU tar from bsdtar from busybox, and zip -v for its compiled-in methods. Anything that will not work on this machine is greyed out with the reason. Unpacking a .zip wants unzip, which is a different program from the zip used to write one, so it is checked on its own.

Members are stored relative to the deepest folder that contains every tagged path, which is shown on the confirm screen. Tag ~/a/b and ~/c and the archive holds a/b and c.

• Commands are run directly, not in a shell. Filenames containing spaces, quotes, $, ; or a leading - are safe.
• Cancelling kills the whole process group, so the compressor goes too. A cancelled pack removes its half-written archive; a cancelled unpack leaves the files it had already extracted, since deleting a folder is not rakpak's call to make.
• Folder sizes are counted in the background. A folder too large to finish counting in a few seconds shows ≥, meaning the real size is at least that.

ruby test/test_rakpak.rb