Object that isn't fetched doesn't pull associations from cache

[elado]

(Using master branch)

# first time - no cache at all
> Product.first.fetch_main_category
  Product Load (0.2ms)  SELECT products.* FROM products LIMIT 1
  Category Load (0.2ms)  SELECT categories.* FROM categories WHERE categories.id = 1 LIMIT 1
 => #<Category id: 1 name: "x">

# second time, not using #fetch - no cache for fetch_main_category
> Product.first.fetch_main_category
  Product Load (0.3ms)  SELECT products.* FROM products LIMIT 1
  Category Load (0.3ms)  SELECT categories.* FROM categories WHERE categories.id = 1 LIMIT 1
 => #<Category id: 1 name: "x">

# using #fetch - no cache for fetch_main_category
> Product.fetch(1).fetch_main_category
  Product Load (0.3ms)  SELECT products.* FROM products WHERE products.id = 1 LIMIT 1
  Category Load (0.2ms)  SELECT categories.* FROM categories WHERE categories.id IN (1)
 => #<Category id: 1 name: "x">

# using #fetch - everything's cached
> Product.fetch(1).fetch_main_category
 => #<Category id: 1 name: "x">

Product class has

belongs_to :main_category, class_name: 'Category'
cache_has_one :main_category, inverse_name: :products

Expected: even when Product.first is called, the calls of fetch_main_category will get it from DB, put the Category object in cache and next calls of Product.first.fetch_main_category will read it from cache.

[bdotdub]

Bump.

This seems weird to me as well. Are the methods only applied to objects coming back from memcache?

[elado]

I have a temporary workaround

class Product < AR::Base
  ...
  def fetch_main_category
    Category.fetch(self.main_category_id)
  end
end

ugly, but works.

[dylanahsmith]

cache_has_one embeds the associated main_category record in the Product's cache entry. It isn't meant to be a shortcut to Category.fetch(self.main_category_id) here.

Why aren't you just using cache_belongs_to? It will do what you seem to be expecting.

[elado]

It's not documented on the README. Will check, thanks!

[dylanahsmith]

You're right, that is a valid bug in the documentation at least.

[dylanahsmith]

cache_belongs_to has been documented, so this doesn't demonstrate a bug in the code.