URL-Encoded Form Support

[MATATAT]

Added support for URL encoded forms to framework. This was based off a quick extension trait that I was using in a project I'm working on.

Alternatively, I wrote a separate version of this that follows the httpmock interface more closely. It was a lot more invasive than this change. Here is a link to the diff if you'd like to consider that approach as well.

[declark1]

Thanks for the contribution @MATATAT! I'll have a look today (both versions) and provide feedback later.

FYI, the DCO check is failing as our organization requires commits to be signed. I think git rebase HEAD~3 --signoff should fix it.

[declark1]

Thanks @MATATAT!

Since form data is body content, this can be simplified to be an additional When body convenience method for the existing BodyMatcher, instead of adding a new matcher and FormBody struct.

I would suggest the following implementation for When (accepting multiple kv-pairs for convenience):

/// Body convenience methods.
impl When {
    // ...

    /// URL encoded form body.
    pub fn form(
        self,
        params: impl IntoIterator<Item = (impl Into<String>, impl Into<String>)>,
    ) -> Self {
        let params = params
            .into_iter()
            .map(|(key, value)| (key.into(), value.into()))
            .collect::<Vec<(String, String)>>();
        let body = serde_urlencoded::to_string(params).unwrap();
        self.push(matchers::body(Body::bytes(body)));
        self
    }
}

Usage:

    let mut mocks = MockSet::new();
    mocks.mock(|when, then| {
        when.post()
            .header("content-type", "application/x-www-form-urlencoded")
            .form([("hello", "world")]);
        then.text("hello world");
    });

I tested and this appears to work, let me know what you think.

[MATATAT]

Thanks @declark1, let me take a look at this. I believe I had something like this before and there was an issue where the matcher didn't end up being consistent since the HashMap doesn't enforce insertion order. So when the body matcher would compare the byte order was sporadically inconsistent.

I'll look back through my commits when I was doing this via an extension trait and see if I left any comments about this (or if my implementation was different).

[MATATAT]

Hey! Sorry it took me a while to circle back around on this.

The problem I was running into (and was able to repro this with your suggested changes) are that using the Body matcher the values must maintain their insert order. Since reqwest is also mapping a HashMap this can get out of order fairly easily. Also, with your proposed input pattern, you can pass a HashMap or HashSet into the When builder, which will cause this to fail.

I do like your approach to the inputs, so I'd like to incorporate that, but I do think there will need to be a separate matcher for the form body.

I might have some time this weekend so I'll make some changes and push another commit.