Upload large content & callbacks

[edsiper]

Duda stack should support an API to allow the server to listen for large data uploads, where Monkey server after receive the HTTP request headers (on PUT/POST), let Duda handle each incoming data frame.

The API should be something like this:

map->static_upload_stream("/upload", int (*callback)(struct duda_request *dr,  void* buffer, ssize_t len));

so all POST/PUT request with content length > 0 that arrives over /upload URI, will be handled by the callback function which also sets a buffer and a size, that function will be invoked many times until the upload is complete or for some reason the connection is closed.

[nalbeza]

Steps to implement feature

This comment is a work in progress.

Monkey

• Modify HTTP parser to allow stream parsing (maybe use something like Ragel ?)

  Ex:

  parser->request_line_cb = ...;
  parser->header_read_cb = ...;
  void parse_input_data(parser_ctx_t* parser, session_request* sr, void* data, size_t len);

  Or just wait for block end and parse whole header block at once

• Add necessary API hooks (placeholder names)

  • on_headers_ready
    Could be used to authenticate request before body is sent (fail early), or setup body streaming
    Must expose an API to enable body chunk callback
    Return value actions:
    • Abort connection, error response
    • Continue, read request body
  • on_body_chunk_read
    Receives read body chunks.
    Return value actions:
    • Abort connection, error response
    • Chunk handled, buffered data can be discarded
    • Chunk handled, continue to buffer data

• Find all code that expects whole request buffered and work around that

Duda