Editorial: less-confusing content-type manipulation algorithm for send()

[domenic]

I spent a lot of time staring at send(), both in reviewing #176 and in implementing this in jsdom. I think the algorithm is too confusing mainly because it doesn't make it clear that the purpose is setting and manipulating the Content-Type header.

Thus, I propose this replacement for the current algorithm. I've written it in pseudo-JS since that was easier for me to rearrange things:

if (body !== null) {
  let extractedMIMEType = null;

  switch (body) {
    case Document:
      request.body = body, serialized, converted to unicode, and utf-8 encoded
    case BodyInit:
      { extractedBody, extractedMIMEType } = extractBody(body);
      request.body = extractedBody;
  }

  if (authorRequestHeaders.contains("Content-Type")) {
    if (body is Document or body is string) {
      let originalAuthorContentType = authorRequestHeaders.get("Content-Type");
      let contentTypeRecord = MIMEType.parse(contentTypeRecord);
      if (contentTypeRecord !== failure && contentTypeRecord.parameters.has("charset")) {
        contentTypeRecord.parameters.set("charset", "UTF-8");
        let newContentTypeSerialized = contentTypeRecord.serialize();
        authorRequestHeaders.set("Content-Type", newContentTypeSerialized);
      }
    }
  } else {
    if (body is HTML document) {
      authorRequestHeaders.set("Content-Type", "text/html;charset=UTF-8");
    }
    if (body is XML document) {
      authorRequestHeaders.set("Content-Type", "application/xml;charset=UTF-8");
    }
    if (extractedMIMEType !== null) {
      authorRequestHeaders.set("Content-Type", extractedMIMEType);
    }
  }
}

The main idea of this rewrite is:

• Separate out setting the request body from setting the Content-Type. Don't concern ourselves with encodings or content types when setting the request body, except to the extent we're forced to by the extract algorithm.
• Make it clear that we're taking different paths depending on whether the author request headers contain Content-Type or not
• Make it clear under what conditions we override the author's Content-Type, and how. Each distinct possible result (parse/reserialize; set to HTML; set to XML; set to extracted MIME type) is clearly a peer of each other, instead of setting up all the variables ahead of time then sometimes using them.

WDYT? I can submit a PR after things settle down, if you like this.

[annevk]

Seems reasonable.