--- /dev/null
+# This module allows you to easily cache all responses from the backend - thus
+# speeding up the I18n aspects of your application quite a bit.
+#
+# To enable caching you can simply include the Cache module to the Simple
+# backend - or whatever other backend you are using:
+#
+# I18n::Backend::Simple.send(:include, I18n::Backend::Cache)
+#
+# You will also need to set a cache store implementation that you want to use:
+#
+# I18n.cache_store = ActiveSupport::Cache.lookup_store(:memory_store)
+#
+# You can use any cache implementation you want that provides the same API as
+# ActiveSupport::Cache (only the methods #fetch and #write are being used).
+#
+# The cache_key implementation assumes that you only pass values to
+# I18n.translate that return a valid key from #hash (see
+# http://www.ruby-doc.org/core/classes/Object.html#M000337).
+#
+# If you use a lambda as a default value in your translation like this:
+#
+# I18n.t(:"date.order", :default => lambda {[:month, :day, :year]})
+#
+# Then you will always have a cache miss, because each time this method
+# is called the lambda will have a different hash value. If you know
+# the result of the lambda is a constant as in the example above, then
+# to cache this you can make the lambda a constant, like this:
+#
+# DEFAULT_DATE_ORDER = lambda {[:month, :day, :year]}
+# ...
+# I18n.t(:"date.order", :default => DEFAULT_DATE_ORDER)
+#
+# If the lambda may result in different values for each call then consider
+# also using the Memoize backend.
+#
+module I18n
+ class << self
+ @@cache_store = nil
+ @@cache_namespace = nil
+
+ def cache_store
+ @@cache_store
+ end
+
+ def cache_store=(store)
+ @@cache_store = store
+ end
+
+ def cache_namespace
+ @@cache_namespace
+ end
+
+ def cache_namespace=(namespace)
+ @@cache_namespace = namespace
+ end
+
+ def perform_caching?
+ !cache_store.nil?
+ end
+ end
+
+ module Backend
+ # TODO Should the cache be cleared if new translations are stored?
+ module Cache
+ def translate(locale, key, options = {})
+ I18n.perform_caching? ? fetch(cache_key(locale, key, options)) { super } : super
+ end
+
+ protected
+
+ def fetch(cache_key, &block)
+ result = _fetch(cache_key, &block)
+ throw(:exception, result) if result.is_a?(MissingTranslation)
+ result = result.dup if result.frozen? rescue result
+ result
+ end
+
+ def _fetch(cache_key, &block)
+ result = I18n.cache_store.read(cache_key) and return result
+ result = catch(:exception, &block)
+ I18n.cache_store.write(cache_key, result) unless result.is_a?(Proc)
+ result
+ end
+
+ def cache_key(locale, key, options)
+ # This assumes that only simple, native Ruby values are passed to I18n.translate.
+ "i18n/#{I18n.cache_namespace}/#{locale}/#{key.hash}/#{USE_INSPECT_HASH ? options.inspect.hash : options.hash}"
+ end
+
+ private
+ # In Ruby < 1.9 the following is true: { :foo => 1, :bar => 2 }.hash == { :foo => 2, :bar => 1 }.hash
+ # Therefore we must use the hash of the inspect string instead to avoid cache key colisions.
+ USE_INSPECT_HASH = RUBY_VERSION <= "1.9"
+ end
+ end
+end