require 'fortitude/rendering_context' if defined?(ActiveSupport) ActiveSupport.on_load(:before_initialize) do ActiveSupport.on_load(:action_view) do require "fortitude/rails/template_handler" end end end module Fortitude class << self def refine_rails_helpers(on_or_off = :not_specified) @refine_rails_helpers = !! on_or_off unless on_or_off == :not_specified !! @refine_rails_helpers end end refine_rails_helpers true end module Fortitude module Rails class Railtie < ::Rails::Railtie config.after_initialize do if Fortitude.refine_rails_helpers require 'fortitude/rails/helpers' Fortitude::Rails::Helpers.apply_refined_helpers_to!(Fortitude::Widget) end if ::Rails.env.development? ::Fortitude::Widget.class_eval do format_output true start_and_end_comments true debug true end end end initializer :fortitude, :before => :set_autoload_paths do |app| # All of this code is involved in setting up autoload_paths to work with Fortitude. # Why so hard? # # We're trying to do something that ActiveSupport::Dependencies -- which is what Rails uses for # class autoloading -- doesn't really support. We want app/views to be on the autoload path, # because there are now Ruby classes living there. (It usually isn't just because all that's there # are template source files, not actual Ruby code.) That isn't an issue, though -- adding it # is trivial (just do # ActiveSupport::Dependencies.autoload_paths << File.join(Rails.root, 'app/views')). # # The real issue is that we want the class app/views/foo/bar.rb to define a class called # Views::Foo::Bar, not just plain Foo::Bar. This is what's different from what # ActiveSupport::Dependencies normally supports; it expects the filesystem path underneath the # root to be exactly identical to the fully-qualified class name. # # Why are we doing this crazy thing? Because we want you to be able to have a view called # app/views/user/password.rb, and _not_ have that conflict with a module you just happen to define # elsewhere called User::Password. If we don't prefix view classes with anything at all, then the # potential for conflicts is enormous. # # As such, we have this code. We'll walk through it step-by-step; note that at the end we *do* # add app/views/ to the autoload path, so all this code is doing is just dealing with the fact that # the fully-qualified classname (Views::Foo::Bar) has one extra component on the front of it # (Views::) when compared to the subpath (foo/bar.rb) underneath what's on the autoload # path (app/views). # Go compute our views root. views_root = File.expand_path(File.join(::Rails.root, 'app', 'views')) # Now, do all this work inside ::ActiveSupport::Dependencies... ::ActiveSupport::Dependencies.module_eval do @@_fortitude_views_root = views_root # This is the method that gets called to auto-generate namespacing empty # modules (_e.g._, the toplevel Views:: module) for directories # under an autoload path. # # The original method says: # # "Does the provided path_suffix correspond to an autoloadable module? # Instead of returning a boolean, the autoload base for this module is # returned." # # So, we just need to strip off the leading +views/+ from the +path_suffix+, # and see if that maps to a directory underneath app/views/; if so, # we'll return the path to .../app/views/. Otherwise, we just # delegate back to the superclass method. def autoloadable_module_with_fortitude?(path_suffix) if path_suffix =~ %r{^views(/.*)?$}i # If we got here, then we were passed a subpath of views/.... subpath = $1 if subpath.blank? || File.directory?(File.join(@@_fortitude_views_root, subpath)) return @@_fortitude_views_root end end with_fortitude_views_removed_from_autoload_path do autoloadable_module_without_fortitude?(path_suffix) end end alias_method_chain :autoloadable_module?, :fortitude # When we delegate back to original methods, we want them to act as if # app/views/ is _not_ on the autoload path. In order to be thread-safe # about that, we couple this method with our override of the writer side of the # mattr_accessor :autoload_paths, which simply prefers the thread-local # that we set to the actual underlying variable. def with_fortitude_views_removed_from_autoload_path begin Thread.current[:_fortitude_autoload_paths_override] = autoload_paths - [ @@_fortitude_views_root ] yield ensure Thread.current[:_fortitude_autoload_paths_override] = nil end end # The use of 'class_eval' here may seem funny, and I think it is, but, without it, # the +@@autoload_paths+ gets interpreted as a class variable for this *Railtie*, # rather than for ::ActiveSupport::Dependencies. (Why is that? Got me...) class_eval <<-EOS def self.autoload_paths Thread.current[:_fortitude_autoload_paths_override] || @@autoload_paths end EOS # The original method says: # # "Search for a file in autoload_paths matching the provided suffix." # # So, we just look to see if the given +path_suffix+ is specifying something like # views/foo/bar; if so, we glue it together properly, removing the initial # views/ first. (Otherwise, the mechanism would expect # Views::Foo::Bar to show up in app/views/views/foo/bar (yes, a double # +views+), since app/views is on the autoload path.) def search_for_file_with_fortitude(path_suffix) # This just makes sure our path always ends in exactly one ".rb", whether it started # with one or not. new_path_suffix = path_suffix.sub(/(\.rb)?$/, ".rb") if new_path_suffix =~ %r{^views(/.*)$}i path = File.join(@@_fortitude_views_root, $1) return path if File.file?(path) end # Make sure that we remove the views autoload path before letting the rest of # the dependency mechanism go searching for files, or else app/views/foo/bar.rb # *will* be found when looking for just ::Foo::Bar. with_fortitude_views_removed_from_autoload_path { search_for_file_without_fortitude(path_suffix) } end alias_method_chain :search_for_file, :fortitude end # And, finally, this is where we add our root to the set of autoload paths. ::ActiveSupport::Dependencies.autoload_paths << views_root # This is our support for partials. Fortitude doesn't really have a distinction between # partials and "full" templates -- everything is just a widget, which is much more elegant -- # but we still want you to be able to render a widget Views::Foo::Bar by saying # render :partial => 'foo/bar' (from ERb, although you can do it from Fortitude if # you want for some reason, too). # # Normally, ActionView only looks for partials in files starting with an underscore. We # do want to allow this, too (in the above case, if you define the widget in the file # app/views/foo/_bar.rb, it will still work fine); however, we also want to allow # you to define it in a file that does _not_ start with an underscore ('cause these are # Ruby classes, and that's just plain weird). # # So, we patch #find_templates: if it's looking for a partial, doesn't find one, and is # searching Fortitude templates (the +.rb+ handler), then we try again, turning off the # +partial+ flag, and return that instead. ::ActionView::PathResolver.class_eval do def find_templates_with_fortitude(name, prefix, partial, details) templates = find_templates_without_fortitude(name, prefix, partial, details) if partial && templates.empty? && details[:handlers] && details[:handlers].include?(:rb) templates = find_templates_without_fortitude(name, prefix, false, details.merge(:handlers => [ :rb ])) end templates end alias_method_chain :find_templates, :fortitude end require "fortitude/rails/template_handler" # This is our support for render :widget. Although, originally, it looked like creating a new subclass # of ActionView::Template was going to be the correct thing to do here, it turns out it isn't: the entire # template system is predicated around the idea that you have a template, which is compiled to output # Ruby source code, and then that gets evaluated to actually generate output. # # Because render :widget => ... takes an already-instantiated widget as input, this simply isn't # possible: you can't reverse-engineer an arbitrary Ruby object into source code, and, without source code, # the whole templating paradigm doesn't make sense. # # So, instead, we simply transform render :widget => ... into a render :text => ... of the # widget's output, and let Rails take it away from there. ::ActionController::Base.class_eval do def fortitude_rendering_context(options) @_fortitude_rendering_context ||= create_fortitude_rendering_context(options) end def create_fortitude_rendering_context(options) ::Fortitude::RenderingContext.new(options) end def render_with_fortitude(*args, &block) if (options = args[0]).kind_of?(Hash) if (widget = options[:widget]) rendering_context = fortitude_rendering_context(:delegate_object => self) widget.render_to(rendering_context) options = options.dup options[:text] = rendering_context.output_buffer_holder.output_buffer.html_safe options[:layout] = true unless options.has_key?(:layout) new_args = [ options ] + args[1..-1] return render_without_fortitude(*new_args, &block) elsif (widget_block = options[:inline]) && (options[:type] == :fortitude) options.delete(:inline) rendering_context = fortitude_rendering_context(:delegate_object => self) widget_class = Class.new(Fortitude::Widgets::Html5) widget_class.use_instance_variables_for_assigns(true) widget_class.extra_assigns(:use) widget_class.send(:define_method, :content, &widget_block) assigns = { } instance_variables.each do |ivar_name| value = instance_variable_get(ivar_name) assigns[$1.to_sym] = value if ivar_name =~ /^@(.*)$/ end assigns = assigns.merge(options[:locals] || { }) widget = widget_class.new(assigns) widget.render_to(rendering_context) options = options.dup options[:text] = rendering_context.output_buffer_holder.output_buffer.html_safe options[:layout] = true unless options.has_key?(:layout) new_args = [ options ] + args[1..-1] return render_without_fortitude(*new_args, &block) end end return render_without_fortitude(*args, &block) end alias_method_chain :render, :fortitude end end end end end