I recently had an issue with use_inline_resources.
This feature's documentation is lackluster, and I learnt about its magic thanks to some scraps of information.
I wanted to share some of my info about how it actually works.

The basics

use_inline_resources is used to separate a LWRP's run context from the main run, making it run in an isolated "mini-run".
Assume the following structure inside cookbook slasher:

  • resources/default.rb:

    # resource definition for slasher LWRP
    actions :swing
    default_action :swing
  • providers/default.rb:

    # provider for slasher LWRP
    action :swing do
      execute 'echo swish'

This causes the following effects:

  • You cannot subscribe to / notify resources that are not part of this resource's context.
    This will break:

    # provider for slasher LWRP
    action :swing do
      execute 'echo swish' do
        notifies :restart,'service[nginx]' # not defined in this LWRP!
  • The external resource is automatically "updated" (meaning it triggers other resources using subscribe/notify) whenever one of the resources in the run changes.
    This will work:

    # External usage of slash
    service nginx
    slash do
      notifies :restart,'service[nginx]' # Will happen because the execute resource will "update"

This behaviour is the encouraged way to implement LWRPs, because it helps encapsulation - they feel like "atomic" resources when used in recipes, and have no need to consider other resources when managing naming, notifications etc., especially other instances of the same resource. It'll probably be enabled by default in Chef13.

The internals

This is the interesting part - how it's actually implemented. If you'll look at the Chef source (not in master, where I guess everything changed because it's the default behaviour), under file lib/chef/provider/lwrp_base.rb, we'll see that the method use_inline_resources does some mixin magic:

def use_inline_resources
    extend InlineResources::ClassMethods
    include InlineResources

The mixed-in module (in the same file) overrides the method action (used to define actions in LWRP) with the following code:

def action(name, &block)
  define_method("action_#{name}") do

As you can see, it calls recipe_eval_with_update_check (instead of instance_eval in the regular version), which will temporarily replace the run context with a new one (with an empty resource collection), run the resource's action, and then inspect the resources defined in this temporary context and see if any of them were updated. If so, it'll mark this resource (the LWRP) as updated too:
(Comments are mine)

def recipe_eval_with_update_check(&block)
  saved_run_context = @run_context                                # Save that for later
  temp_run_context = @run_context.dup                             # Create new run context
  @run_context = temp_run_context                                 # Set it as current
  @run_context.resource_collection = Chef::ResourceCollection.new # New and empty collection

  return_value = instance_eval(&block)                            # Compile LWRP
  Chef::Runner.new(@run_context).converge                         # Execute compiled LWRP
  return_value                                                    # Return result
ensure                                                            # Always happens, even if we had an error
  @run_context = saved_run_context                                # Restore old run context
  if temp_run_context.resource_collection.any? {|r| r.updated? }  # Search for modified resources
    new_resource.updated_by_last_action(true)                     # If found, update LWRP

And that's how it's done!