class Roda module RodaPlugins # The per_thread_caching plugin changes the default cache # from being a shared thread safe cache to a separate cache per # thread. This means getting or setting values no longer # needs a mutex on non-MRI ruby implementations, which may be # faster when using a thread pool. However, since the caches # are no longer shared, this will take up more memory. # # Note that it does not make sense to use this plugin on MRI, # since the default cache on MRI doesn't use a mutex as it # is already thread safe due to the GVL. # # Using this plugin changes the matcher regexp cache to use # per-thread caches, and changes the default for future # thread-safe caches to use per-thread caches. # # If you want the render plugin's template cache to use # per-thread caches, you should load this plugin before the # render plugin. module PerThreadCaching def self.configure(app) app::RodaRequest.match_pattern_cache = app.thread_safe_cache end class Cache # Mutex used to ensure multiple per-thread caches # don't use the same key MUTEX = ::Mutex.new n = 0 # Auto incrementing number proc used to make sure # multiple thread-thread caches don't use the same key. N = lambda{MUTEX.synchronize{n += 1}} # Store unique symbol used to look up in the per # thread caches. def initialize @o = :"roda_per_thread_cache_#{N.call}" end # Return the current thread's cached value. def [](key) _hash[key] end # Set the current thread's cached value. def []=(key, value) _hash[key] = value end private # The current thread's cache. def _hash ::Thread.current[@o] ||= {} end end module ClassMethods # Use the per-thread cache instead of the default cache. def thread_safe_cache Cache.new end end end register_plugin(:per_thread_caching, PerThreadCaching) end end