feat(headers): use a Vec + a map for Cookie

[dorfsmay]

Use a Vec and a map for the Cookie header in order to satisfy the need
for both duplicate names (keys) from a client perspective and single
value for a given key from a server perspective .

BREAKING CHANGE: This change the format of the cookie header, any
previous implementation will have to be changed.

• The commit messages match the guidelines in https://github.com/hyperium/hyper/blob/master/CONTRIBUTING.md#git-commit-guidelines

[dorfsmay]

@seanmonstar @LegNeato

As part of the larger discussion in #1145

This is a first stab to make sure that's going in the right direction.

• It doesn't matter how you dice it, at the end of the day, a "Cookie header" is really a string representation of a Vec<(String, String)>, so let's use that, and use methods and/or additional elements in the structure to provide the functionalities we need, such as a get() returning the first value for a given key.
• I know HashMap, I'll re-submit a different PR with VecMap, right now, this compiles and demonstrates the concepts.
• I use a Vec<(String, String)> + a HashMap (VecMap later), in order to be able to both not lose duplicate keys, and only give the value for the first key on a et
• Another way to do the above would be to just use a Vec<(String, String)>, and make the get() method walk it. It would be cheaper in memory, but more expensive if an implementer looks up more than one key.
• Using owned value in the HasMap is a waste of memory, but I am very green to Rust (weeks), I wasted hours trying to make it use reference to the owned values inside the Vec and make it work in hyper... Again this is more to show what I think would be the right implementation for the Cookie header. Also, we could roll with this as it gets the signatures right, and implement internal references later if it is worth it (and maybe somebody more experience add that!).
• I added a set, diff from a push (append) as per @LegNeato sugestion
• Still missing: a proper iterator. I wasted another few hours on this, will work more on that later.

[GitCop]

Thanks for contributing! Unfortunately, I'm here to tell you there were the following style issues with your Pull Request:

• Commit: 6c74b5f
• Commits must be in the following format: %{type}(%{scope}): %{description}

Guidelines are available at https://github.com/hyperium/hyper/blob/master/CONTRIBUTING.md

---

This message was auto-generated by https://gitcop.com

[seanmonstar]

Thanks for exploring this with me. I believe the code in #1152 may be in a slightly better place, but we wouldn't have gotten there without your efforts to bring up the question and discuss how to solve it!

[seanmonstar]

My understanding of the suggestion for this method was to clear the values of only this key. This would seem confusing:

let mut cookie = Cookie::new();
cookie.set("foo", "bar");
cookie.set("baz", "quux");
// where did my foo=bar go!
assert_eq!(cookie.to_string(), "baz=quux");

[dorfsmay]

I understood the suggestion as push()/append() to add an extra cookie name/value to the header, and set() to for the entire header to this one value.

[seanmonstar]

This would allocate a brand new String, before then writing it to the formatter. It can be done without the temporary allocated, such as write!(f, "{}={}", key, val).

[dorfsmay]

Great point. Thanks.

[seanmonstar]

The VecMap is sufficient on it's own. There no need to double the strings here. The VecMap has insert and append, and iter gets all the key-value pairs, with duplicates.

Also, a Cow<'static, str> can be used, which allows someone who wants to set the same key and value on all cookies to do so without allocating and copying memory for it:

let mut cookie = Cookie::new();
// no allocation or copy required
cookie.set("foo", "bar");
// dynamic string, use a String
cookie.set(nonce_key.to_string(), nonce_val.to_string());

[seanmonstar]

We'd still want a Debug implementation. It can use formatter.debug_map().

[dorfsmay]

Agreed, but with my implementation, I would have had to add a Debug trait to VecMap. I wanted to get the code out fast for sake of discussion, rather than spend too many hours on an implementation that wasn't going to be used.

[dorfsmay]

Thanks for taking the time to go through my code and the comments, this is really helpful.