Refactored HttpClient, HttpServer and WebsocketConnection.

[slaff]

• Better Code Design (There were some places that deserved to be called "big-ball-of-mud" ).
• Faster and powerful HttpParser used in both HttpClient and HttpServer.
• Support for pipelining in both HttpClient and HttpServer.
• Support for connection reuse for both HttpClient and HttpServer.
• SSL session resumption support for HttpClient and HttpServer.
• Added ResourceTree to HttpServer to allow more flexible definition of resources.
• Added streaming support to both Http and WebSocket processing.
• rBootHttpUpdate should stop now on failure as early as possible.
• Changed the ContentType code to allow easier definition of new mime types.

Backwards-Incompatible changes

• WebSocket is renamed to WebSocketConnection to reflect better its meaning and intended use.
• Removed the tightly coupled websocket methods inside the HttpServer. You can use WebSocketResource and add it to the HttpServer resource tree to achieve the same results.
• Moved writeInit(), writeFlash(const u8 *data, u16 size) and writeEnd() methods from rBootHttpUpdate to rBootItemOutputStream.
• TemplateFileStream::setVarsFromRequest needs to be refactored completely. The current code couples TemplateFileStream to HttpRequest, which is too restrictive. TemplateFileStream::setVars method should be introduced that accepts HashMap<String,String>
• HttpRequest::getRequestMethod() is removed. Use HttpRequest::method instead.
• CommandExecutor with WebsocketConnection won't work.

[Vity01]

God bless the guy (@slaff) who rewrote the HttpClient. 👍
It was the weakest part of the Sming. I can't wait for 3.3.0.

[Vity01]

@slaff Do you close/clean/free input stream for body in HttpRequest somewhere or the caller is responsible?

[slaff]

@Vity01 If you use the HttpClient then at the end of an HTTP request the request data is freed in HttpConnection::staticOnMessageComplete and similar thing is happening for the HttpServer in HttpServerConnection::staticOnMessageComplete.

[Vity01]

@slaff thanks for the info, maybe the line comment might be helpful also for other developers
Thanks again for the great job.

[lucas-bremond]

Great job!

Just out of curiosity, what is the rationale behind:

aRequest.getRequestMethod() == RequestMethod::GET

being changed into:

aRequest.method == HTTP_GET

?

It's shorter indeed, but seems more brittle to me.
Was it performance driven?