Transparently explore all archive contents.
This repository is in early prototyping state, and it is NON-FUNCTIONAL. This was the result of one rainy afternoon of coding, and it is not usable yet.
I wanted to have a FUSE filesystem that transparently decompressed any archives on-the-fly, so I could navigate the archive contents using any tool, just as if those archives were directories.
In other words, I wanted this:
$ ls
Before/
After/
$ ls Before
Photos.zip
Report.zip
$ ls After
(empty)
$ fuse-archive-explorer Before After
("Before" directory is now mounted onto "After".)
$ ls After
Photos.zip
Photos_zip/
Report.zip
Report_zip/
$ ls After/Photos_zip/
DSC01234.JPG
DSC01235.JPG
$ mupdf After/Report_zip/2023-Q1/foobar.pdf
(opens the PDF from inside that zipfile)
- A directory subtree should be accessible from the mountpoint (i.e. loopback or passthrough).
- Archive files (e.g.
*.zip) would also be accessible as a directory (e.g. by replacing.zipwith_zip).- Undefined behavior if there is already a file or directory with the same name as the new zip-as-directory fake name. This is so rare in the real world that I don't want to bother with it.
- Read-only access.
Optional requirements:
- Support for multiple archive types.
- ZIP should be supported.
- Compressed tarballs may be supported.
- 7z may as well be supported.
- RAR may be supported, but the licensing is so troublesome that is likely not worth the effort.
- Support for additional extensions for the same kind of archive.
- ZIP files can be recognized as
*.zipbut also*.cbzand others. - Tarballs can be
*.tar,*.tar.gz,*.tgz, etc.
- ZIP files can be recognized as
- Support for detecting archives based on the file signature.
- May be quite slow, as it requires reading a few bytes on all files.
- Support for non-archive types.
- It could be nice to be able to extract data from non-archive files. Data that aren't real files.
- Metadata from multimedia files (audio/video/image).
- PNG chunks from a PNG file.
- Still frames from animated GIFs or from animated APNG.
- Each icon size from a
.icofile. - Streams from a video file (e.g. individual audio tracks, individual subtitles).
- This is just nice-to-have, and definitely beyond the initial purpose of this repository. But still very cool to have.
- Nested archives may be supported.
- If
foo.zipcontainsbar.zip, it would be nice to access the contents through./foo_zip/bar_zip/contents. - This is rare enough in real-world that I wouldn't bother implementing it at the beginning.
- There is a large potential for CPU an RAM bottlenecks.
- If
- Write support is an welcome addition.
- Having write support for the un-archived files would be simple enough to implement and convenient. This way, extracting a single file from inside an archive would be as simple as
cp foo.zip/bar.txt . - Having full write support for the archived files seems like trouble. A lot of work needed, and terrible performance.
- Having write support for the un-archived files would be simple enough to implement and convenient. This way, extracting a single file from inside an archive would be as simple as
This is in a very very early stage. I started coding, I can list fake directory entries, and that's all. I haven't implemented anything else.
I certainly need more time to think on how to implement it correctly, and even more time to implement it and make sure it works fine.
For simplicity, I started coding it in Python, using python-fuse. I knew this would be less performant than other languages, but I just wanted to get something done quickly enough. This was supposed to be fun hacky project, not a production one. (Too bad the Python bindings for FUSE are not well maintained.)
This project is not really usable. It's a starting point, and not much else. If you are unsure if you want to try it, then don't. But if you like the idea and want a starting point to start implementing your own FUSE filesystem, then go ahead and take a look at this project. But it's likely you will find better examples elsewhere anyway.