EM-Synchrony
Collection of convenience classes and primitives to help untangle evented code, plus a number of patched EM clients to make them Fiber aware. To learn more, please see: Untangling Evented Code with Ruby Fibers.
- Fiber aware ConnectionPool with sync/async query support
- Fiber aware Iterator to allow concurrency control & mixing of sync / async
- Fiber aware async inline support: turns any async function into sync
- Fiber aware Multi-request interface for any callback enabled clients
- Fiber aware TCPSocket replacement, powered by EventMachine
- Fiber aware Thread, Mutex, ConditionVariable clases
- Fiber aware sleep, defer, system
Supported clients:
- mysql2: .query is synchronous, while .aquery is async (see specs)
- activerecord: require synchrony/activerecord, set your AR adapter to em_mysql2 and you should be good to go
- em-http-request: .get, etc are synchronous, while .aget, etc are async
- em-memcached & remcached: .get, etc, and .multi_* methods are synchronous
- em-mongo: .find, .first are synchronous
- mongoid: all functions synchronous, plus Rails compatibility
- em-jack: a[method]'s are async, and all regular jack method's are synchronous
- AMQP: most of functions are synchronous (see specs)
Other clients with native Fiber support:
- redis: contains synchrony code right within the driver
- synchrony also supports em-redis and em-hiredis (see specs), but unless you specifically need either of those, use the official redis gem
Fiber-aware Iterator: mixing sync / async code
Allows you to perform each, map, inject on a collection of any asynchronous tasks. To advance the iterator, simply call iter.next, or iter.return(result). The iterator will not exit until you advance through the entire collection. Additionally, you can specify the desired concurrency level! Ex: crawling a web-site, but you want to have at most 5 connections open at any one time.
require "em-synchrony"
require "em-synchrony/em-http"
EM.synchrony do
concurrency = 2
urls = ['http://url.1.com', 'http://url2.com']
# iterator will execute async blocks until completion, .each, .inject also work!
results = EM::Synchrony::Iterator.new(urls, concurrency).map do |url, iter|
# fire async requests, on completion advance the iterator
http = EventMachine::HttpRequest.new(url).aget
http.callback { iter.return(http) }
http.errback { iter.return(http) }
end
p results # all completed requests
EventMachine.stop
end
Or, you can use FiberIterator to hide the async nature of em-http:
require "em-synchrony"
require "em-synchrony/em-http"
require "em-synchrony/fiber_iterator"
EM.synchrony do
concurrency = 2
urls = ['http://url.1.com', 'http://url2.com']
results = []
EM::Synchrony::FiberIterator.new(urls, concurrency).each do |url|
resp = EventMachine::HttpRequest.new(url).get
results.push resp.response
end
p results # all completed requests
EventMachine.stop
end
Fiber-aware ConnectionPool shared by a fiber:
Allows you to create a pool of resources which are then shared by one or more fibers. A good example is a collection of long-lived database connections. The pool will automatically block and wake up the fibers as the connections become available.
require "em-synchrony"
require "em-synchrony/mysql2"
EventMachine.synchrony do
db = EventMachine::Synchrony::ConnectionPool.new(size: 2) do
Mysql2::EM::Client.new
end
multi = EventMachine::Synchrony::Multi.new
multi.add :a, db.aquery("select sleep(1)")
multi.add :b, db.aquery("select sleep(1)")
res = multi.perform
p "Look ma, no callbacks, and parallel MySQL requests!"
p res
EventMachine.stop
end
Fiber-aware Multi interface: parallel HTTP requests
Allows you to fire simultaneous requests and wait for all of them to complete (success or error) before advancing. Concurrently fetching many HTTP pages at once is a good example; parallel SQL queries is another. Technically, this functionality can be also achieved by using the Synchrony Iterator shown above.
require "em-synchrony"
require "em-synchrony/em-http"
EventMachine.synchrony do
multi = EventMachine::Synchrony::Multi.new
multi.add :a, EventMachine::HttpRequest.new("http://www.postrank.com").aget
multi.add :b, EventMachine::HttpRequest.new("http://www.postrank.com").apost
res = multi.perform
p "Look ma, no callbacks, and parallel HTTP requests!"
p res
EventMachine.stop
end
Fiber-aware & EventMachine backed TCPSocket:
This is dangerous territory - you've been warned. You can patch your base TCPSocket class to make any/all libraries depending on TCPSocket be actually powered by EventMachine and Fibers under the hood.
require "em-synchrony"
require "lib/em-synchrony"
require "net/http"
EM.synchrony do
# replace default Socket code to use EventMachine Sockets instead
TCPSocket = EventMachine::Synchrony::TCPSocket
Net::HTTP.get_print 'www.google.com', '/index.html'
EM.stop
end
Inline synchronization & Fiber sleep:
Allows you to inline/synchronize any callback interface to behave as if it was a blocking call. Simply pass any callback object to Synchrony.sync and it will do the right thing: the fiber will be resumed once the callback/errback fires. Likewise, use Synchrony.sleep to avoid blocking the main thread if you need to put one of your workers to sleep.
require "em-synchrony"
require "em-synchrony/em-http"
EM.synchrony do
# pass a callback enabled client to sync to automatically resume it when callback fires
result = EM::Synchrony.sync EventMachine::HttpRequest.new('http://www.gooogle.com/').aget
p result
# pause execution for 2 seconds
EM::Synchrony.sleep(2)
EM.stop
end
Async ActiveRecord:
Allows you to use async ActiveRecord within Rails and outside of Rails (see async-rails). If you need to control the connection pool size, use rack/fiber_pool.
require "em-synchrony"
require "em-synchrony/mysql2"
require "em-synchrony/activerecord"
ActiveRecord::Base.establish_connection(
:adapter => 'em_mysql2',
:database => 'widgets'
)
result = Widget.all.to_a
Hooks
em-synchrony already provides fiber-aware calls for sleep, system and defer. When mixing fiber-aware code with other gems, these might use non-fiber-aware versions which result in unexpected behavior: calling sleep
would pause the whole reactor instead of
a single fiber. For that reason, hooks into the Kernel are provided to override the default behavior to e.g. add logging or redirect these calls their fiber-aware versions.
# Adding logging but still executes the actual sleep
require "em-synchrony"
log = Logger.new(STDOUT)
EM::Synchrony.on_sleep do |*args|
log.warn "Kernel.sleep called by:"
caller.each { |line| log.warn line }
sleep(*args) # Calls the actual sleep
end
# Redirects to EM::Synchrony.sleep
require "em-synchrony"
log = Logger.new(STDOUT)
EM::Synchrony.on_sleep do |*args|
EM::Synchrony.sleep(*args)
end
License
The MIT License - Copyright (c) 2011 Ilya Grigorik