explosive.fuse 0.5

Mount archives as a filesystem.

Introduction

This package, explosive.fuse, provides a command-line tool, explode, for mounting of archive files (such as zip) to a directory using FUSE, enabling access to individual file entries within the compressed archives as normal files directly without the typical intermediate step of decompression to some temporary location, i.e. all file decompression is done dynamically, on demand. This library is constructed with modularity in mind to enable customization, such as how the file entries within the archives are to be presented at the mounted location, and that the helper classes and functions within can be reused elsewhere.

Currently, the following archive formats are supported:

• zip (builtin package zipfile)

• rar (require third-party package unrar)

Archive formats that require third-party packages are not enabled by default. To enable, please install the packages as specified, and more information about them can be accessed via the links provided above.

Installation

If all system level dependencies are satisfied, the installation process is simply a single pip call, like so:

$ pip install explosive.fuse

This can be done either at the system level (i.e. using sudo) to make this available for all users on the system, or inside a virtualenv for just a single user.

System level dependencies

ExplosiveFUSE requires the FUSE kernel module for your operating system and its user space tools be available, and Python 2.7/3.3+/PyPy.

First off, install the FUSE kernel module and user space tools if they are not already installed.

Linux

Arch Linux:

$ sudo pacman -S fuse

Debian/Ubuntu:

$ sudo apt-get install fuse

Gentoo (also, please refer to its wiki entry on FUSE):

$ sudo emerge --ask sys-fs/fuse

RedHat/CentOS/Fedora:

$ sudo yum install fuse

OS X

This has been somewhat tested through the experimental build features provided by Travis-CI, with dependencies installed via Homebrew. For installation of Homebrew, please refer to instructions outlined at http://brew.sh/.

Once that’s done, the packages for FUSE for OS X and pyenv (which provides Python version management) can be installed using the following commands:

$ brew update
$ brew install caskroom/cask/osxfuse
$ brew install pyenv
$ pyenv install 3.5.2  # or your desired version of Python
$ pyenv global 3.5.2
$ pyenv rehash

Other installation methods

This may be installed into the default user specific library directory (typically ~/.local) using the following command:

$ pip install --user explosive.fuse

The executable will then be available at ~/.local/bin/explode.

Or be installed in a virtualenv, too:

$ virtualenv ~/.venv
$ source ~/.venv/bin/activate
$ pip install explosive.fuse

Naturally, installation from source can be done, and provided git is installed, the latest development version can be done like so:

$ pip install git+https://github.com/metatoaster/explosive.fuse#egg=explosive.fuse

Alternatively, manual usage of git and python setup.py develop will work also.

Usage

Simply invoking explode from the shell this help message will be presented:

usage: explode [-h] [-l <strategy>] [--layout-info] [-d] [-f] [-m]
               [--manager-dir [MANAGER_DIR]] [--overwrite] [--omit-arcname]
               [-V]
               dir archives [archives ...]

In the most simple form the command can simply be invoked like so:

$ mkdir /tmp/mnt
$ explode /tmp/mnt demo1.zip

This will mount the contents of demo1.zip to /tmp/mnt. To verify that this worked, a simple ls can be used:

$ ls -l /tmp/mnt/
total 0
-r--r--r-- 1 user user 33 Oct 26 23:19 file1
-r--r--r-- 1 user user 33 Oct 26 23:19 file2
-r--r--r-- 1 user user 33 Oct 26 23:19 file3
-r--r--r-- 1 user user 33 Oct 26 23:19 file4
-r--r--r-- 1 user user 33 Oct 26 23:19 file5
-r--r--r-- 1 user user 33 Oct 26 23:19 file6

Files are presented as being owned by the user that created this mount point. For specifics on access permissions, please consult the fuse user manual (i.e. man fuse).

To unmount, simply call:

$ fusermount -u /tmp/mnt/

Or terminate the process if it was ran in the foreground.

It is possible to explode multiple archives onto the target directory:

$ explode /tmp/mnt demo1.zip demo2.zip

By default, a new layout strategy will be used, which will include the name of the source archive file. This can be verified:

$ ls -l /tmp/mnt/
total 0
dr-xr-xr-x 2 user user 0 Oct 26 23:22 demo1.zip
dr-xr-xr-x 2 user user 0 Oct 26 23:22 demo2.zip

Layout Strategies

The way the file entries are laid out in the resulting filesystem can be modified by the use of a layout strategy. This is specified using the -l or the --layout flag. Naturally, the final result is also influenced by the usage of the --overwrite and the --omit-arcname flags and the arguments associated with each of the strategies (which are specified by appending :, followed by the value of each positional argument(s)). Detailed information on every available strategies are available by calling explode --layout-info, but for completeness sake the following strategies are provided by a default installation:

codepage

Decode the filename entries into unicode from the specified codepage. Example: -l codepage:shift_jis will decode filenames that look like é▒é±é╔é┐é═ into こんにちは.

default

Present file entries as they were within their respective directory structures to the root of its source archive.

flatten

Flattens the directory structure to the root of the mount point by replacing all path separators for each file entries with the _ character by default. This character can be specified by using the argument syntax (e.g. use -l flatten:- will replace all path separators with the - character.)

junk

Junk paths, keep only directories counting from root up to the level specified for a positive keep number, otherwise junking all but the absolute number of keep levels previous to the basename of the filename for a negative keep number. Default is to keep no directories. Useful value is 1 if it is desirable to keep the source archive name as a container directory (i.e. -l junk:1) if --omit-arcname is not used.

An important note: by default, the basename of the archive file will be prepended to each of its file entries before being filtered through the layout strategy, unless the --omit-arcname flag is used.

Flags for fine-tuning filesystem behavior

--debug

Print debug messages to stdout.

--foreground

Run in foreground.

-m, --manager

Enable the symlink manager directory. This option exposes all the archive files under the management directory (defined by the --manager-dir flag, default is .manager under the root of the mount point) as symlinks. Creating symlinks to valid archive files will add the file entries in them to the filesystem, and removing the symlinks will remove its associated entries from the filesystem.

--omit-arcname

Sometimes it may be desirable to omit the name of the source archive files from the generated paths.

For example, if we have multiple archive files with names SNS_001.zip up to SNS_100.zip, and inside there we simply have files like 01.jpg up to 20.jpg lying at the root level, activating the --omit-arcname flag flag will result in only 20 files from SNS_001.zip archive being accessible as by default as that was the first file specified to be loaded.

-s, --splitext-arcname

Sometimes it may be desirable to split the filename extenxion out of the name of the source archive files from the generated paths.

--overwrite

Useful when there are multiple file entries of the same name from multiple archives and only the latest one is desired, this flag will “overwrite” any existing entries the mapping process may encounter.

Troubleshooting

Error messages

Mounting shows the following error message:

fusermount: failed to open /etc/fuse.conf: Permission denied

This can be safely ignored, or alternatively have your system’s administrator grant you read access to that file by putting your account into the fuse user group or equivalent on your system, or change the permission to that file to world readable, as that file does not contain any sensitive information under typical usage.

Other issues

If you encountered any other problems using this software please file an issue using the issue tracker for this project.

License

This work is licensed under GNU Generic Public License, version 3.

Changelog

0.5 (2018-07-13)

• Support for dropping of the file name extension (mostly to improve the presentation of filename sorting for certain file managers); may be toggled through the -s or --splitext-arcname flag.

• Be able to extract zip files that do not have a .zip file name extension, as this is now the default fallback to better support the exploding of common file formats that make use of zip, e.g. odt or docx.

0.4 (2016-08-20)

• RAR archive format support added using the package unrar.

0.3 (2015-12-12)

• Mappings now use absolute path to ensure this works in daemon mode.

• New layout strategy: codepage can be used to remap non-unicode name encodings to unicode.

• File ownership now shown as being owned by the user that started the mount process.

• Symlink management support added; this is enabled using the -m flag, optionally with --manager-dir to explicitly change where that lies. This allows loading and unloading of files via the adding and removing of symlinks in the management directory.

• Decreased memory consumption and performance from reads as inflate is called only on demand. This however requires single-thread mode for the mean time.

0.2 (2015-10-31)

• Removed specialized archive layout strategies (i.e. the zip* ones)

• By default, basename of archive will be prepended to the path for each file entries. This behavior can directly be disabled using the --omit-arcname flag. (Naturally, a layout strategy can still modify that path, like junk.

• All pathmaker callables are now actually factories that produce the actual pathmaker. They can now take arguments to influence how the produced pathmaker behaves. The arguments can be entered using the extended -l or --layout syntax. Details documented in --layout-info.

• Implemented the --overwrite flag, allowing newer entries to “overwrite” existing ones.

• A number of other internal API changes.

0.1 (2015-10-26)

• Initial release with just basic zip file support.

• File extraction is naive, such that the entire contents of a single given file entry will be extracted to memory per read.