Another usability issue with map_err about error-chain

[Yamakaky]

Do you want to try a PR?

[golddranks]

So, I tried and made two additional APIs. My work is here: golddranks@76d5ccf

I added a CausedErr trait that is analogous to ChainErr, but works directly on types implementing the std::error::Error trait, not Result. With that, I was able to get the ugly

e => Err::<(), diesel::result::Error>(e).chain_err(|| "Error when trying to retrieve user!").unwrap_err(),

row to

e => e.caused_err(|| "Error when trying to retrieve user!"),

I could also refactor the implementation of ChainErr to use that. (Just basic DRY)

I also added a .to_err() helper method for ErrorKind, which works exactly like .into() but causes less problems with type inference. (Basically it's but a sugar to Error::from(), but I like the conciseness, method syntax and lack of :: better.)

Still WIP in the sense that I haven't touched the docs. But what do you think about these changes? Do they seem worthwile, @brson ? Shall I send a PR?

[Yamakaky]

It seems close to cause() from https://github.com/tailhook/quick-error?
Edit: not really...

[Yamakaky]

@golddranks please review #62

[TedDriggs]

I see the PR for this has stalled due to an inference problem; are there any plans to revive it?

FWIW, I hit the exact same issue (also when using map_err) and found that this was sufficient to get moving again:

impl Error {
    pub fn caused<K: Into<ErrorKind>>(self, kind: K) -> Self {
        Error::with_chain(self, kind)
    }
}

Would a simple approach like that work for this?

[TedDriggs]

Since #141 was merged, should this be closed?

[golddranks]

Indeed! Thanks!