module Kgio
See the README
Constants
- LOCALHOST
The IPv4 address of UNIX domain sockets, useful for creating Rack (and CGI) servers that also serve HTTP traffic over UNIX domain sockets.
- MSG_FASTOPEN
- SOCK_CLOEXEC
Maps to the SOCK_CLOEXEC constant in Linux for setting the close-on-exec flag on newly accepted descriptors. This is enabled by default, and there is usually no reason to disable close-on-exec for accepted sockets.
- SOCK_NONBLOCK
Maps to the SOCK_NONBLOCK constant in Linux for setting the non-blocking flag on newly accepted sockets. This is usually unnecessary as sockets are made non-blocking whenever non-blocking methods are used.
- TCP_FASTOPEN
- WaitReadable
Kgio::PipeMethods#kgio_tryread and Kgio::SocketMethods#kgio_tryread will return :wait_readable when waiting for a read is required.
- WaitWritable
Kgio::PipeMethods#kgio_trywrite and Kgio::SocketMethods#kgio_trywrite will return :wait_writable when waiting for a read is required.
Public Class Methods
Returns the default class for newly accepted sockets when kgio_accept or kgio_tryaccept are not passed arguments
static VALUE get_accepted(VALUE klass) { return cClientSocket; }
Sets the default class for newly accepted sockets. This is legacy behavior, kgio_accept and kgio_tryaccept now take optional class arguments to override this value.
static VALUE set_accepted(VALUE klass, VALUE aclass) { VALUE tmp; if (NIL_P(aclass)) aclass = cKgio_Socket; tmp = rb_funcall(aclass, rb_intern("included_modules"), 0, 0); tmp = rb_funcall(tmp, rb_intern("include?"), 1, mSocketMethods); if (tmp != Qtrue) rb_raise(rb_eTypeError, "class must include Kgio::SocketMethods"); cClientSocket = aclass; return aclass; }
Sets whether or not Kgio::Socket objects created by Kgio::TCPServer#kgio_accept, Kgio::TCPServer#kgio_tryaccept, Kgio::UNIXServer#kgio_accept, and Kgio::UNIXServer#kgio_tryaccept default to being created with the FD_CLOEXEC file descriptor flag.
This is on by default, as there is little reason to deal to enable it for client sockets on a socket server.
Deprecated, use the per-socket flags for kgio_*accept instead.
static VALUE set_cloexec(VALUE mod, VALUE boolean) { switch (TYPE(boolean)) { case T_TRUE: accept4_flags |= SOCK_CLOEXEC; return boolean; case T_FALSE: accept4_flags &= ~SOCK_CLOEXEC; return boolean; } rb_raise(rb_eTypeError, "not true or false"); return Qnil; }
Returns true if newly accepted Kgio::Sockets are created with the FD_CLOEXEC file descriptor flag, false if not.
Deprecated, use the per-socket flags for kgio_*accept instead.
static VALUE get_cloexec(VALUE mod) { return (accept4_flags & SOCK_CLOEXEC) == SOCK_CLOEXEC ? Qtrue : Qfalse; }
Sets whether or not Kgio::Socket objects created by Kgio::TCPServer#kgio_accept, Kgio::TCPServer#kgio_tryaccept, Kgio::UNIXServer#kgio_accept, and Kgio::UNIXServer#kgio_tryaccept are created with the O_NONBLOCK file status flag.
This defaults to false
for GNU/Linux where MSG_DONTWAIT is
available (and on newer GNU/Linux, accept4() may also set the non-blocking
flag. This defaults to true
on non-GNU/Linux systems.
This is always true on Ruby implementations using user-space threads.
Deprecated, use the per-socket flags for kgio_*accept instead.
static VALUE set_nonblock(VALUE mod, VALUE boolean) { switch (TYPE(boolean)) { case T_TRUE: accept4_flags |= SOCK_NONBLOCK; return boolean; case T_FALSE: accept4_flags &= ~SOCK_NONBLOCK; return boolean; } rb_raise(rb_eTypeError, "not true or false"); return Qnil; }
Returns true if newly accepted Kgio::Sockets are created with the O_NONBLOCK file status flag, false if not.
Deprecated, use the per-socket flags for kgio_*accept instead.
static VALUE get_nonblock(VALUE mod) { return (accept4_flags & SOCK_NONBLOCK)==SOCK_NONBLOCK ? Qtrue : Qfalse; }
Enables or disables autopush for sockets created with kgio_accept and kgio_tryaccept methods. Autopush relies on TCP_CORK/TCP_NOPUSH being enabled on the listen socket.
Only available on systems with TCP_CORK (Linux) or TCP_NOPUSH (FreeBSD, and maybe other *BSDs).
static VALUE s_set_autopush(VALUE self, VALUE val) { enabled = RTEST(val); return val; }
Returns whether or not autopush is enabled.
Only available on systems with TCP_CORK (Linux) or TCP_NOPUSH (FreeBSD, and maybe other *BSDs).
static VALUE s_get_autopush(VALUE self) { return enabled ? Qtrue : Qfalse; }
Accepts an input hash with IO objects to wait for as the key and the events
to wait for as its value. The events may either be
:wait_readable
or :wait_writable
symbols or a
Fixnum mask of Kgio::POLL* constants:
Kgio::POLLIN - there is data to read Kgio::POLLPRI - there is urgent data to read Kgio::POLLOUT - writing will not block Kgio::POLLRDHUP - peer has shutdown writes (Linux 2.6.17+ only)
Timeout is specified in Integer milliseconds just like the underlying poll(2), not in seconds like IO.select. A nil timeout means to wait forever. It must be an Integer or nil.
::poll modifies and returns its input hash on success with the IO-like object as the key and an Integer mask of events as the hash value. It can return any of the events specified in the input above, along with the following events:
Kgio::POLLERR - error condition occurred on the descriptor Kgio::POLLHUP - hang up Kgio::POLLNVAL - invalid request (bad file descriptor)
This method is only available under Ruby 1.9 or any other implementations that uses native threads and rb_thread_blocking_region()
static VALUE s_poll(int argc, VALUE *argv, VALUE self) { VALUE timeout; struct poll_args a; rb_scan_args(argc, argv, "11", &a.ios, &timeout); a.timeout = num2timeout(timeout); a.fds = NULL; a.fd_to_io = NULL; return rb_ensure(do_poll, (VALUE)&a, poll_free, (VALUE)&a); }
Like ::tryread, except it uses MSG_PEEK so it does not drain the socket buffer. This can only be used on sockets and not pipe objects. Maybe used in place of Kgio::SocketMethods#kgio_trypeek for non-Kgio objects
static VALUE s_trypeek(int argc, VALUE *argv, VALUE mod) { if (argc <= 1) rb_raise(rb_eArgError, "wrong number of arguments"); return my_peek(0, argc - 1, &argv[1], argv[0]); }
Returns nil on EOF. Returns :wait_readable if EAGAIN is encountered.
Maybe used in place of Kgio::PipeMethods#kgio_tryread for non-Kgio objects
static VALUE s_tryread(int argc, VALUE *argv, VALUE mod) { if (argc <= 1) rb_raise(rb_eArgError, "wrong number of arguments"); return my_read(0, argc - 1, &argv[1], argv[0]); }
Returns nil if the write was completed in full.
Returns a String containing the unwritten portion if EAGAIN was encountered, but some portion was successfully written.
Returns :wait_writable if EAGAIN is encountered and nothing was written.
Maybe used in place of Kgio::PipeMethods#kgio_trywrite for non-Kgio objects
static VALUE s_trywrite(VALUE mod, VALUE io, VALUE str) { return my_write(io, str, 0); }
Returns nil if the write was completed in full.
Returns a Array of strings containing the unwritten portion if EAGAIN was encountered, but some portion was successfully written.
Returns :wait_writable if EAGAIN is encountered and nothing was written.
Maybe used in place of Kgio::PipeMethods#kgio_trywritev for non-Kgio objects
static VALUE s_trywritev(VALUE mod, VALUE io, VALUE ary) { return kgio_trywritev(io, ary); }