Content-based filetype detection

[scop]

It seems file type detection is based on only the file (base)name and its extension.

Would be nice to improve on that some and peek into the file content, and take into account for example

• Shebang, if any
• File magic
• Emacs style -*- something -*- embedded configs
• Vi and friends configs like 'ex: filetype=sh(ditto forvi:` at least)

I imagine something like https://github.com/ebassi/xdg-mime-rs could be useful at least for the two first items in the above list.

[epage]

Concerns for each of these:

• Translating what we find to our standard term for a file type
• Deciding precedence
• Whether to allow the user to extend it and how

Shebang, if any

This will be important for cases like rust-lang/rfcs#3424

Specifically for shebangs, we'd need to decide how to parse the various cases (whether /usr/bin/env is used, flags passed to it, its in a different path, etc)

[epage]

Forgot to add, xdg-mime-rs seems like it would be platform specific (loading a mime DB that won't be on all platforms) while we need to be cross-platform.

We also need to weigh out the costs of this since this can lead to extra IO