Add support for TAR file handling in EPUB processing

[andrestobelem]

While I was reading Modern Software Engineering: Doing What Works to Build Better Software Faster by Dave Farley, I ran into a small but annoying issue that felt very familiar.

What happened

I was working with a bunch of .epub files and, like most people, I assumed they would all follow the usual rule: an EPUB is just a ZIP file with a specific structure inside.

Turns out that wasn’t always true.

Some of the files were actually TAR archives (sometimes compressed), just renamed to .epub. As expected, standard EPUB libraries couldn’t read them at all.

There were a couple of extra complications too:

• Non-standard directory layouts: the book contents were inside nested folders instead of being at the root, which broke relative paths.
• System files mixed in: hidden OS metadata and resource files showed up in the archive and occasionally caused validation or parsing issues.

What I ended up doing

Rather than special-casing everything later, I added a small pre-processing step to clean things up before reading the files:

• Check the real file type
  Instead of trusting the file extension, the code looks at the file signature to see what it actually is.

• Convert in memory
  If the file is a TAR, it gets converted to a ZIP on the fly, using in-memory buffers only.

• Figure out the actual root
  The required mimetype file is used as an anchor to find the real root of the EPUB, and any extra container directories are stripped out.

• Keep the rest simple
  After that, the rest of the system always works with a normal ZIP-based EPUB, no matter how the file was originally packaged.

Nothing fancy, but it made the whole flow a bit more resilient and avoided a bunch of edge cases later on.