upgrade httpclient and include http-cookie

[hfaulds]

We discussed this in #113 and #105

To summarise. At the moment our Rets::HttpClient#http_cookie implementation assumes the httpclient is using WebAgent::CookieManager and WebAgent::Cookie under the hood.

If a project requires httpclient 2.6 and http-cookie then this breaks as httpclient loads HttpClient::CookieManager and a different cookie implementation.

A couple quick fixes would be to:

• explicitly require 'httpclient/webagent-cookie' in our Rets::HttpClient
• explicitly depend on http-cookie and change out Rets::HttpClient#http_cookie

Moving away from WebAgent::CookieManager is a step towards being able to store cookies in storage other than the filesystem (#105). It is more up to date with RFC and more flexible as well.

Rets::CookieManager exists to expose jar publicly. I made a PR to httpclient so that in the future we should be able to avoid this: nahi/httpclient#254.

This will also make it easy to provide a custom cookie jar implementation.

[hfaulds]

This whole class is a stop-gap for nahi/httpclient#254.

I was a bit loathe to write our own cookie manager just to inject and expose the cookie jar (I'm a bit confused about what parts of HTTPClient rely on what cookie implementations as it seems like internally it uses WebAgent::Cookie some places).

[hfaulds]

This needs to also be a new version of the Gem as it changes dependencies.

[hfaulds]

Did some testing against real rets servers. http-cookie's cookie jar expects the cookie store file to be present and exceptions if a non-existant file is provided. We could handle this either in our code or force people to create the files themselves, I'm open to suggestions.

[hfaulds]

This also spams log output with Cookie#domain returns dot-less domain name now. Use Cookie#dot_domain if you need "." at the beginning.

httpclient appears to set up both the warn and then also call this internally

[dougcole]

code looks fine. am I right that this would supersede #113?

As long as you've tested it, I'm game to ship it.

[summera]

@hfaulds @dougcole Maybe we can combine efforts on this and #113 by merging our PRs together? I think both have ideas we can combine and we are stepping on each others toes at this point.

I just pushed an update to #113 that cleans up the tests and uses http-cookie as we discussed.

For this, do you think it may be premature to override the cookie manager and expose the cookie jar? I see that this would make sense when #105 develops into something, but does not seem necessary at the moment. Also, may be good to have the dependency on http-cookie be ~> 1.0.0 so we don't have another compatibility issue. Same goes for dependency on httpclient.

[hfaulds]

Closing in favour of #113. Might end up with some of this logic if we write a custom cookie store but this goes beyond what it was supposed to.