Add PathManager class to cache interim paths (#7732)

Merge pull request 7732

lib/jekyll.rb

@@ -65,6 +65,7 @@ module Jekyll
   autoload :LogAdapter,          "jekyll/log_adapter"
   autoload :Page,                "jekyll/page"
   autoload :PageWithoutAFile,    "jekyll/page_without_a_file"
+  autoload :PathManager,         "jekyll/path_manager"
   autoload :PluginManager,       "jekyll/plugin_manager"
   autoload :Publisher,           "jekyll/publisher"
   autoload :Reader,              "jekyll/reader"

lib/jekyll/entry_filter.rb

@@ -90,12 +90,12 @@ module Jekyll
     # Check if an entry matches a specific pattern.
     # Returns true if path matches against any glob pattern, else false.
     def glob_include?(enumerator, entry)
-      entry_with_source = File.join(site.source, entry)
+      entry_with_source = PathManager.join(site.source, entry)
 
       enumerator.any? do |pattern|
         case pattern
         when String
-          pattern_with_source = File.join(site.source, pattern)
+          pattern_with_source = PathManager.join(site.source, pattern)
 
           File.fnmatch?(pattern_with_source, entry_with_source) ||
             entry_with_source.start_with?(pattern_with_source)

lib/jekyll/page.rb

@@ -47,7 +47,7 @@ module Jekyll
               end
 
       process(name)
-      read_yaml(File.join(base, dir), name)
+      read_yaml(PathManager.join(base, dir), name)
 
       data.default_proc = proc do |_, key|
         site.frontmatter_defaults.find(relative_path, type, key)

lib/jekyll/path_manager.rb

@@ -0,0 +1,31 @@
+# frozen_string_literal: true
+
+module Jekyll
+  # A singleton class that caches frozen instances of path strings returned from its methods.
+  #
+  # NOTE:
+  #   This class exists because `File.join` allocates an Array and returns a new String on every
+  #   call using **the same arguments**. Caching the result means reduced memory usage.
+  #   However, the caches are never flushed so that they can be used even when a site is
+  #   regenerating. The results are frozen to deter mutation of the cached string.
+  #
+  #   Therefore, employ this class only for situations where caching the result is necessary
+  #   for performance reasons.
+  #
+  class PathManager
+    # This class cannot be initialized from outside
+    private_class_method :new
+
+    # Wraps `File.join` to cache the frozen result.
+    # Reassigns `nil`, empty strings and empty arrays to a frozen empty string beforehand.
+    #
+    # Returns a frozen string.
+    def self.join(base, item)
+      base = "" if base.nil? || base.empty?
+      item = "" if item.nil? || item.empty?
+      @join ||= {}
+      @join[base] ||= {}
+      @join[base][item] ||= File.join(base, item).freeze
+    end
+  end
+end

lib/jekyll/reader.rb

@@ -85,7 +85,7 @@ module Jekyll
     def retrieve_dirs(_base, dir, dot_dirs)
       dot_dirs.each do |file|
         dir_path = site.in_source_dir(dir, file)
-        rel_path = File.join(dir, file)
+        rel_path = PathManager.join(dir, file)
         @site.reader.read_directories(rel_path) unless @site.dest.chomp("/") == dir_path
       end
     end

lib/jekyll/tags/include.rb

@@ -100,7 +100,7 @@ module Jekyll
       def locate_include_file(context, file, safe)
         includes_dirs = tag_includes_dirs(context)
         includes_dirs.each do |dir|
-          path = File.join(dir.to_s, file.to_s)
+          path = PathManager.join(dir, file)
           return path if valid_include_file?(path, dir.to_s, safe)
         end
         raise IOError, could_not_locate_message(file, includes_dirs, safe)