From 24403498ba8582d29f55aab16ffd5920dec1c669 Mon Sep 17 00:00:00 2001
From: Nicholas Seckar <nseckar@gmail.com>
Date: Mon, 27 Mar 2006 03:59:35 +0000
Subject: Add CachingTools::HashCaching to simplify the creation of nested,
 autofilling hashes.

git-svn-id: http://svn-commit.rubyonrails.org/rails/trunk@4059 5ecf4fe2-1ee6-0310-87b1-e25e094e27de
---
 activesupport/lib/active_support/caching_tools.rb | 62 +++++++++++++++++++++++
 1 file changed, 62 insertions(+)
 create mode 100644 activesupport/lib/active_support/caching_tools.rb

(limited to 'activesupport/lib/active_support')

diff --git a/activesupport/lib/active_support/caching_tools.rb b/activesupport/lib/active_support/caching_tools.rb
new file mode 100644
index 0000000000..c889c148b4
--- /dev/null
+++ b/activesupport/lib/active_support/caching_tools.rb
@@ -0,0 +1,62 @@
+module ActiveSupport
+  module CachingTools #:nodoc:
+    
+    # Provide shortcuts to simply the creation of nested default hashes. This
+    # pattern is useful, common practice, and unsightly when done manually.
+    module HashCaching
+      # Dynamically create a nested hash structure used to cache calls to +method_name+
+      # The cache method is named +#{method_name}_cache+ unless :as => :alternate_name
+      # is given.
+      #
+      # The hash structure is created using nested Hash.new. For example:
+      # 
+      #   def slow_method(a, b) a ** b end
+      # 
+      # can be cached using hash_cache :slow_method, which will define the method
+      # slow_method_cache. We can then find the result of a ** b using:
+      # 
+      #   slow_method_cache[a][b]
+      # 
+      # The hash structure returned by slow_method_cache would look like this:
+      # 
+      #   Hash.new do |as, a|
+      #     as[a] = Hash.new do |bs, b|
+      #       bs[b] = slow_method(a, b)
+      #     end
+      #   end
+      # 
+      # The generated code is actually compressed onto a single line to maintain
+      # sensible backtrace signatures.
+      #
+      def hash_cache(method_name, options = {})
+        selector = options[:as] || "#{method_name}_cache"
+        method = self.instance_method(method_name)
+        
+        args = []
+        code = "def #{selector}(); @#{selector} ||= "
+        
+        (1..method.arity).each do |n|
+          args << "v#{n}"
+          code << "Hash.new {|h#{n}, v#{n}| h#{n}[v#{n}] = "
+        end
+        
+        # Add the method call with arguments, followed by closing braces and end.
+        code << "#{method_name}(#{args * ', '}) #{'}' * method.arity} end"
+        
+        # Extract the line number information from the caller. Exceptions arising
+        # in the generated code should point to the +hash_cache :...+ line.
+        if caller[0] && /^(.*):(\d+)$/ =~ caller[0]
+          file, line_number = $1, $2.to_i
+        else # We can't give good trackback info; fallback to this line:
+          file, line_number = __FILE__, __LINE__
+        end
+        
+        # We use eval rather than building proc's because it allows us to avoid
+        # linking the Hash's to this method's binding. Experience has shown that
+        # doing so can cause obtuse memory leaks.
+        class_eval code, file, line_number
+      end
+    end
+    
+  end
+end
-- 
cgit v1.2.3