Module Uq_io

module Uq_io: sig .. end

Unified engines for stream I/O

type in_buffer

type out_buffer

Buffers that can be attached to a device to get buffered I/O

type in_device = [ `Async_in of Uq_engines.async_in_channel * Unixqueue.event_system
       | `Buffer_in of in_buffer
       | `Count_in of (int -> unit) * in_device
       | `Multiplex of Uq_engines.multiplex_controller
       | `Polldescr of Netsys.fd_style * Unix.file_descr * Unixqueue.event_system ]

Currently supported devices for input:

`Polldescr(st,fd,esys): The poll system call is used with file descriptor fd to wait for incoming data. The event system esys is the underlying event queue. This works well for pipes, sockets etc. but not for normal files. The style st can be obtained from fd via Netsys.get_fd_style.
`Multiplex mplex: The multiplex controller mplex is used as device.
`Buffer buf: Data comes from the buffer buf (which in turn is connected with a second device)
`Count_in(f,d): Data is read from d, and every time a few bytes n are read the function f n is called (which may raise an exception)

Generally, it is not well supported to read in parallel several times from the same device.

type out_device = [ `Async_out of Uq_engines.async_out_channel * Unixqueue.event_system
       | `Buffer_out of out_buffer
       | `Count_out of (int -> unit) * out_device
       | `Multiplex of Uq_engines.multiplex_controller
       | `Polldescr of Netsys.fd_style * Unix.file_descr * Unixqueue.event_system ]

Currently supported devices for output:

`Polldescr(fd,esys): The poll system call is used with file descriptor fd to wait until data can be output. The event system esys is the underlying event queue. This works well for pipes, sockets etc. but not for normal files.
`Multiplex mplex: The multiplex controller mplex is used as device.
`Buffer buf: Data is written to the buffer buf (which in turn is connected with a second device)
`Count_out(f,d): Data is written to d, and every time a few bytes n are written the function f n is called (which may raise an exception)

Generally, it is not well supported to write in parallel several times to the same device.

type in_bdevice = [ `Buffer_in of in_buffer ]

Devices with look-ahead

type string_like = [ `Memory of Netsys_mem.memory | `String of string ]

The user can pass data buffers that base either on strings or on bigarrays of char (memory). Note that `Memory is not supported for all devices or device configurations.

val device_supports_memory : [ `Async_in of Uq_engines.async_in_channel * Unixqueue.event_system
       | `Async_out of Uq_engines.async_out_channel * Unixqueue.event_system
       | `Buffer_in of in_buffer
       | `Buffer_out of out_buffer
       | `Count_in of (int -> unit) * in_device
       | `Count_out of (int -> unit) * out_device
       | `Multiplex of Uq_engines.multiplex_controller
       | `Polldescr of Netsys.fd_style * Unix.file_descr * Unixqueue.event_system ] ->
       bool

Returns whether `Memory buffers are supported

exception Line_too_long

May be raised by Uq_io.input_line_e

Input

val input_e : [< in_device ] ->
       string_like -> int -> int -> int Uq_engines.engine

let e = input_e d s pos len: Reads data from d and puts it into the string s starting at pos and with maximum length len. When data is available, the engine e transitions to `Done n where n is the number of actually read bytes.

If len>0 and no bytes can be read because the end is reached, the engine transitions to `Error End_of_file.

val really_input_e : [< in_device ] ->
       string_like -> int -> int -> unit Uq_engines.engine

let e = input_e d s pos len: Reads data from d and puts it into the string s starting at pos and with length len. Exactly len bytes are read, and when done, the engine e transitions to `Done ().

If the end of the file is reached before len bytes are read, the engine transitions to `Error End_of_file.

val input_line_e : ?max_len:int -> in_bdevice -> string Uq_engines.engine

let e = input_line_e d: Reads the next line from d and transitions to `Done line when done. Note that this is only supported for a buffered device!

If the end of the file is already reached when this function is called, the engine transitions to `Error End_of_file.

If max_len is set, this is the maximum length of the line (including LF). If exceeded, the engine transitions to `Error Line_too_long.

val eof_as_none : 'a Uq_engines.final_state -> 'a option Uq_engines.final_state

Represents EOF as None. Useful in the combination

 input_e d s p l >> eof_as_none

and

 input_line_e d >> eof_as_none

where >> is from Uq_engines.Operators

Output

val output_e : [< out_device ] ->
       string_like -> int -> int -> int Uq_engines.engine

let e = output_e d s pos len: Outputs data to d and takes it from the string s starting at pos and with maximum length len. When data is written, the engine e transitions to `Done n where n is the number of actually written bytes.

val really_output_e : [< out_device ] ->
       string_like -> int -> int -> unit Uq_engines.engine

let e = really_output_e d s pos len: Outputs data to d and takes it from the string s starting at pos and with length len. When all data is written, the engine e transitions to `Done ().

val output_string_e : [< out_device ] -> string -> unit Uq_engines.engine

let e = output_string_e d s: Outputs the string s to d, and transitions to `Done() when done.

val output_memory_e : [< out_device ] -> Netsys_mem.memory -> unit Uq_engines.engine

let e = output_string_e d m: Outputs the bigarray m to d, and transitions to `Done() when done.

val output_netbuffer_e : [< out_device ] -> Netbuffer.t -> unit Uq_engines.engine

let e = output_string_e d b: Outputs the contents of b to d, and transitions to `Done() when done.

val write_eof_e : [< out_device ] -> bool Uq_engines.engine

let e = write_eof_e d: For devices supporting half-open connections, this engine writes the EOF marker and transitions to `Done true. For other devices nothing happens, and the engine transitions to `Done false. (The only way to signal EOF is to shut down the device, see below.)

Note that the effect of write_eof_e cannot be buffered. Because of this, the io_buffer flushes all data first (i.e. write_eof_e implies the effect of flush_e).

val copy_e : ?len:int ->
       ?len64:int64 ->
       [< in_device ] -> [< out_device ] -> int64 Uq_engines.engine

let e = copy_e d_in d_out: Copies data from d_in to d_out, and transitions to `Done n when all data is copied (where n are the number of copied bytes). By default, d_in is read until end of file. If len is passed, at most this number of bytes are copied. The length can also be given as int64 in len64.

val flush_e : [< out_device ] -> unit Uq_engines.engine

let e = flush_e d: If d has an internal buffer, all data is written out to d. If there is no such buffer, this is a no-op. When done, the engine transitions to `Done().

Shutdown

The shutdown is the last part of the protocol. Although it is often done autonomously by the kernel, this interface supports user-implemented shutdowns (e.g. for SSL).

The shutdown can be skipped, and the device can be inactivated immediately. For some devices, the other side of the I/O stream will then see an error, though.

The shutdown is always for both the input and the output circuit of the device.

val shutdown_e : ?linger:float ->
       [< `Async_in of Uq_engines.async_in_channel * Unixqueue.event_system
        | `Async_out of Uq_engines.async_out_channel * Unixqueue.event_system
        | `Buffer_in of in_buffer
        | `Buffer_out of out_buffer
        | `Count_in of (int -> unit) * in_device
        | `Count_out of (int -> unit) * out_device
        | `Multiplex of Uq_engines.multiplex_controller
        | `Polldescr of Netsys.fd_style * Unix.file_descr * Unixqueue.event_system ] ->
       unit Uq_engines.engine

Performs a regular shutdown of the device. The linger argument may be used to configure a non-default linger timeout. The engine transitions to `Done() when done.

The shutdown also releases the OS resources (closes the descriptor etc.), but only if successful.

Note that the effect of shutdown_e cannot be buffered. Because of this, the io_buffer flushes all data first (i.e. shutdown_e implies the effect of flush_e). Input data available in the buffer can still be read after the shutdown.

val inactivate : [< `Async_in of Uq_engines.async_in_channel * Unixqueue.event_system
        | `Async_out of Uq_engines.async_out_channel * Unixqueue.event_system
        | `Buffer_in of in_buffer
        | `Buffer_out of out_buffer
        | `Count_in of (int -> unit) * in_device
        | `Count_out of (int -> unit) * out_device
        | `Multiplex of Uq_engines.multiplex_controller
        | `Polldescr of Netsys.fd_style * Unix.file_descr * Unixqueue.event_system ] ->
       unit

Releases the OS resources immediately. This is the right thing to do when aborting the communication, or for cleanup after an I/O error. It is wrong to inactivate after a successful shutdown, because the shutdown already includes the inactivation.

Buffers

val create_in_buffer : [< in_device ] -> in_buffer

Provides a buffered version of the in_device

val in_buffer_length : in_buffer -> int

The length

val in_buffer_blit : in_buffer -> int -> string_like -> int -> int -> unit

Blit to a string or memory buffer

val in_buffer_fill_e : in_buffer -> bool Uq_engines.engine

Requests that the buffer is filled more than currently, and transitions to `Done eof when there is more data, or the EOF is reached (eof=true).

val create_out_buffer : max:int option -> [< out_device ] -> out_buffer

Provides a buffered version of the out_device. The argument max is the maximum number of bytes to buffer. This can also be set to None meaning no limit.

val filter_out_buffer : max:int option ->
       Netchannels.io_obj_channel -> [< out_device ] -> out_buffer

filter_out_buffer ~max p d: The data written to this device is redirected via pipe p and finally written to d.

This web site is published by Informatikbüro Gerd Stolpmann

Plasma	GitLab	Archive
Projects	Blog	Knowledge