Adding a new OptionExt trait?

[dgrnbrg]

Hello, I'm interested in contributing a new Ext trait for Option, so that I can have chain_err on options when they're None. Often in my code I am working with a vector or iterator that should have a certain number of elements, and I want to bail when that's not the case. Currently, I have tons of option.ok_or(Error::from(format!("problem because of {:?}", context)))? throughout my code, and I think it'd be cool to be able to instead do: option.chain_err(|| format!("problem beause of {:?}", context)). That way, I wouldn't need to compute the error in the success case, and I'd delegate the error casting to type inference. Would you accept this as a PR?

[Yamakaky]

Yes! Though, could you first try to impl ResultExt for Option? I think it should be good.

[dgrnbrg]

Here's my first draft at the feature--I still need to write tests, and try integrating it throughout my code. Does that look good to you WRT to how I used State?

[Yamakaky]

You can use from_kind which creates a default state for you.

[dgrnbrg]

So, you're saying to use $crate::Error::from_kind(callback().into()), rather than making a new ChainedError with default State?

[Yamakaky]

Yes, that's it. It does the same thing: https://brson.github.io/error-chain/src/error_chain/error_chain.rs.html#121-126

[dgrnbrg]

Here's my updated changes: dgrnbrg@d02131f

I'm using from_kind like you suggested, and I've also had to remove the E type from ResultExt, since it turns out ResultExt doesn't ever actually need to be parameterized on E, and it's impossible to determine when chaining with an Option<T>.

I've also included a new method ChainedError::with_boxed_chain, to create a way to start with a boxed Error trait object and get the crate's concrete type of Error with the boxed error as context.

Finally, you can see the commented out code I wrote which would allow for chain_err to be invoked on a boxed error, as I wrote about in #152, if you have any thoughts on how to get the compiler to accept it.

[Yamakaky]

Sorry for the delay.

Seems nice. Could you open a PR?

[dgrnbrg]

I'm closing this in favor of the corresponding PR #156.