Replace some VFS API to use AsRef<Path>

[eirnym]

I have a proposition to replace some VFS API to use AsRef<Path> instead of Path. This would allow to use &str literals, without additional .as_ref().

[SecSamDev]

Sounds good to me. There are still things to review throughout the library, especially ergonomics, to make it easier to use and improve the error system. I'm going to try to get it done this week.

[SecSamDev]

Reviewing this, it's something I thought about at the time. But you can't mix traits with generics so easily:

fn read_to_string<P: AsRef<Path>>(&mut self, path: P) -> ForensicResult<String>;

This will cause all functions that return Box to fail at compile time since the trait cannot be dyn compatible.

The alternative is to create ergonomic wrappers:

impl dyn VirtualFileSystem {
    /// Read the entire contents of a file into a string.
    pub fn read_to_string_path<P: AsRef<Path>>(
        &mut self,
        path: P,
    ) -> ForensicResult<String> {
        self.read_to_string(path.as_ref())
    }
}

[eirnym]

It makes sense. Such VFS wrapper would be a good idea

[eirnym]

About ergonomics of other of your crates. I have some comments, but they're very small and I'd like rather ask why you built API in the way you've built. Sometimes it's better to understand it to make question more precise rather than complain like "why I can't implement this??" Or "I was unable to implement that".

[SecSamDev]

Mainly because all the available tools are too tightly coupled, and the backup storage can't be swapped. In a way, I wanted to make it more flexible and decoupled. Like Lego pieces that can be chained together. It doesn't matter whether the artifacts being processed are inside a ZIP file, your local file system, a remote one, or even a running remote machine (and thats only for the VFS). This way, I can reuse the analysis tools in my own tools and also in other people's.
Additionally, I have a pipeline that processes forensic triages with other random tools, and each tool speaks its own language. The output format is different, and I can't tell if the tool detected errors or not. Even if one has a way to do it, I need to integrate that part manually.

Resume:

• Impossibility of using existing tools in my code, since most are tools, not libraries, and are in Python.
• Each tool speaks its own language: the output format is a mess.
• Parsing stderr and stdout for pipeline integration (this is hell).
• More than once, a tool changed version and broke my pipeline.
• I tried Plaso, but it's too slow when you have an incident with more than 50 triages.