Basic Usage

Usage

LightIO is all about Concurrency and IO.

Use LightIO::Beam to write concurrency code.

result = []
b1 = LightIO::Beam.new {result << 1}
b2 = LightIO::Beam.new {result << 2}
b1.join
b2.join
puts result # 1, 2
puts b1.alive?, b2.alive? # false, false

Beam is a cheaper version thread, use it do the "one thread per client" job, you can gain better concurrency performance.

# echo server example, use telnet to test
LightIO::TCPServer.open('127.0.0.1', 3000) do |server|
  puts "start echo server"
  while (socket = server.accept)
    LightIO::Beam.new(socket){|s| 
      _, port, host = s.peeraddr
      puts "accept connection from #{host}:#{port}"
      loop {s << s.readpartial(4096) rescue break puts("client leave #{host}:#{port}")}
    }
  end
end

Cause beam is not a "real" thread, it will blocking all beams if you do a system IO blocking operation.

LightIO provide IO related classes/modules under LightIO namespace, those modules implemented the same stdlib API(some of them just wrap the stdlib class), use those modules can trigger Beam control-flow transfer, you must use those libraries to instead stdlib to avoid whole beams blocking.

# example: use LightIO::Queue to pass messages between beams
q = LightIO::Queue.new
b1 = LightIO::Beam.new {q.pop}
b2 = LightIO::Beam.new {q << "it will blocking forever if use ::Queue"}
b1.join
b2.join
puts b1.value, b2.value
# it will blocking forever if use ::Queue
# <LightIO::Library::Queue:0x007f818387ef18>

Choose Backend

Cause LightIO use nio4r to provide IO loop, and nio4r support multi backends, LightIO allow user use env variable to choose one.

export LIGHTIO_BACKEND=epoll

# print current backend
puts "available: #{NIO::Selector.backends}"
puts "using: #{LightIO::IOloop.current.backend}"