1
0
Fork 0
forked from OpenNeo/impress

stop caching item preview species images, and fix the bad query instead

Most of the reasoning is documented in the big comment. In short, we tried
to solve the problem with caching, but the caching should hardly be necessary
now that the bottleneck should be fixed. We'll see on production if it
actually solves the whole problem, but I've confirmed in the console that
redefining this function makes random_basic_per_species (as called during
rendering) a ton faster. And this way we keep our randomness, woo!
This commit is contained in:
Emi Matchu 2013-06-23 22:35:27 -07:00
parent d858f33083
commit 3c127569fe
2 changed files with 10 additions and 7 deletions

View file

@ -23,7 +23,15 @@ class PetType < ActiveRecord::Base
lambda { includes({:color => :translations, :species => :translations}) }
def self.standard_pet_types_by_species_id
PetType.where(:color_id => Color.basic).includes_child_translations.
# If we include Color.basic in the query, rather than getting the ID array
# first, it'll do this as a big fat JOIN unnecessarily. On production,
# there are tons of pet types and translations to send down, so tons of
# duplicate data is sent and must be merged together properly, both of
# which take serious time. So, this is a surprisingly significant
# performance boost - though I was surprised and impressed that Rails
# includes relations as subqueries. That's cool.
basic_color_ids = Color.basic.select([:id]).map(&:id)
PetType.where(color_id: basic_color_ids).includes_child_translations.
group_by(&:species_id)
end

View file

@ -79,11 +79,6 @@
#item-preview
#item-preview-species
-# Does this cache ever need to be swept? I don't think so. These days we
-# can identify body-specific items instantly, and, if it's instead
-# particular to a single species, we definitely know that the first time
-# we render.
- localized_cache "items/#{@item.id} item_preview_species" do
= standard_species_images_for(@item)
#item-preview-error
#item-preview-swf= t '.preview.requirements_not_met'