Class type Hydro_transport.hydro_multiplex_controller

class type hydro_multiplex_controller = object .. end

method alive : bool

If the controller is alive, the socket is not yet completely down.

method event_system : Unixqueue.event_system

Returns the event system

method getsockname : sockaddr

The address of this socket

method getpeername : sockaddr

The address of the peer's socket. Only available if the socket is connected. (Fails otherwise.)

method transport_protocol_type : Hydro_types.transport_protocol_type

The transport protocol

method peer_user_name : string option

If the transport mechanism provides a way to authenticate the peer, it can return the name here.

method reading : bool

True iff there is a reader

method read_eof : bool

Whether the EOF marker has been read

method start_reading : ?peek:(unit -> unit) ->
       ?before_record:(int -> sockaddr -> unit) ->
       when_done:((Hydro_types.msg_header * Netbuffer.t * sockaddr)
                  result_eof -> unit) ->
       unit -> unit

Start reading from the connection. When a whole message has been received, the when_done callback is invoked (with header, the full body in the netbuffer, and the socket address of the peer).

This starts one-time read job only, i.e. it is not restarted after when_done has been invoked.

It is an error to start reading several times.

peek: This function is called immediately before a data chunk is read from the underlying data connection.

before_record: If passed, this function is called back whenever a record of data is started. The integer is the estimated size of the message in bytes. It is guaranteed that the function is invoked at least once before when_done.

method skip_message : unit -> unit

Skips the current/next received message, i.e. the when_done callback will not be invoked for it.

method writing : bool

True iff there is a writer

method start_writing : when_done:(unit result -> unit) ->
       Hydro_types.msg_buffer -> sockaddr -> unit

Starts writing the message. Invoked when_done when it is written, or an error condition is reached.

This starts one-time write job only, i.e. it is not restarted after when_done has been invoked.

It is an error to start writing several times.

method start_writing_eof : when_done:(unit result -> unit) -> unit -> unit

method cancel_rd_polling : unit -> unit

Cancels polling for the next input message. This method must not be called from the before_record callback function. Polling can be resumed by calling start_reading again.

method abort_rw : unit -> unit

Aborts the current reader and/or writer forever. Note that there is no clean way of resuming reading and/or writing. The readers/writers are not notified about cancellation.

method start_shutting_down : when_done:(unit result -> unit) -> unit -> unit

Start shutting down the connection. After going through the shutdown procedure, the when_done callback is invoked reporting the success or failure.

The underlying file descriptor (if any) is not closed. A shutdown is only a protocol handshake. After a shutdown,read_eof is true. Call inactivate to close the descriptor.

method cancel_shutting_down : unit -> unit

Cancels the shutdown procedure. After that, the state of the connection is undefined. The when_done callback is invoked with the `Cancelled.

It is no error if no shutdown is in progress.

method set_timeout : notify:(unit -> unit) -> float -> unit

If there is no read or write activity for the passed number of seconds, the notify callback is invoked.

FIXME: Currently, the time out on missing read activity may be overly sensitive. Don't know how to fix.

method inactivate : unit -> unit

Inactivates the connection immediately, and releases any resources the controller is responsible for (e.g. closes file descriptors). Note that this is more than cancelling all pending operations and shutting the connection down. However, the details of this method are implementation-defined. Callbacks are not invoked.