Introduce stash_count(&self) -> Result<usize, Error>

[twe4ked]

Would it be possible to introduce something like this to Repository? I'm not sure if the goals of this project are to keep it mapping to the libgit2 directly or not. I'm also not sure if allowing not requiring a mutable reference here is sound.

/// Count number of git stashes
pub fn stash_count(&self) -> Result<usize, Error> {
    unsafe {
        let mut count = 0;
        let mut data = StashCbData {
            callback: &mut |_, _, _| {
                count += 1;
                true
            },
        };
        let cb: raw::git_stash_cb = Some(stash_cb);
        try_call!(raw::git_stash_foreach(
            self.raw(),
            cb,
            &mut data as *mut _ as *mut _
        ));
        Ok(count)
    }
}

Based of the stash_foreach implementation.

[extrawurst]

I like the train of thought here and I appreciate you took the time to work this out, but I am sceptical about this because:

1. It introduces a helper that is actually not present in the raw libgit2 that we bind to.
2. It increases the surface area of using unsafe functions even though the same can be achieved by using the already existing wrapper around git_stash_foreach namely Repository::stash_foreach docs

---

1. is a decision that has to be made on the repo owners level because it depends on the vision for this crate - right now it looks like a 'safe wrapper' around the bindings to libgit2 and not more which is nice because it basically means whatever method u come across here you can reason about it being in a 1:1 relationship with one c function in the underlying lib

2. independent of 1. I would argue this does not need to be implemented using unsafe.

just my two cents :)

[twe4ked]

1. I thought that might be the case but I didn't see it stated explicitly (in my quick look)
2. It might not need unsafe but ideally we don't need a mutable reference to self which is my main reason for wanting to add another function.

Both of these reasons are why I opened an issue rather than a PR :)

[DanielEScherzer]

No need for counting with a callback, just use the size of the reflog - PR sent