Skip to content

magnusvk/counter_culture

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

counter_culture Build Status

Turbo-charged counter caches for your Rails app. Huge improvements over the Rails standard counter caches:

  • Updates counter cache when values change, not just when creating and destroying
  • Supports counter caches through multiple levels of relations
  • Supports dynamic column names, making it possible to split up the counter cache for different types of objects
  • Can keep a running count, or a running total

Tested against Ruby 2.6, 2.7, 3.0, 3.1, 3.2 and 3.3 and against the latest patch releases of Rails 5.2, 6.0, 6.1, 7.0 and 7.1.

Please note that -- unlike Rails' built-in counter-caches -- counter_culture does not currently change the behavior of the .size method on ActiveRecord associations. If you want to avoid a database query and read the cached value, please use the attribute name containing the counter cache directly.

product.categories.size  # => will lead to a SELECT COUNT(*) query
product.categories_count # => will use counter cache without query

Installation

Add counter_culture to your Gemfile:

gem 'counter_culture', '~> 3.2'

Then run bundle install

Database Schema

You must create the necessary columns for all counter caches. You can use counter_culture's generator to create a skeleton migration:

rails generate counter_culture Category products_count

Which will generate a migration with code like the following:

add_column :categories, :products_count, :integer, null: false, default: 0

Note that the column must be NOT NULL and have a default of zero for this gem to work correctly.

If you are adding counter caches to existing data, you must add code to manually populate their values to the generated migration.

Usage

Simple counter-cache

Has many association

class Product < ActiveRecord::Base
  belongs_to :category
  counter_culture :category
end

class Category < ActiveRecord::Base
  has_many :products
end

Now, the Category model will keep an up-to-date counter-cache in the products_count column of the categories table.

Many to many association

class User < ActiveRecord::Base
  has_many :group_memberships
  has_many :groups, through: :group_memberships
end

class Group < ActiveRecord::Base
  has_many :group_memberships
  has_many :members, through: :group_memberships, class: "User"
end

class GroupMembership < ActiveRecord::Base
  belongs_to :group
  belongs_to :member, class: "User"
  counter_culture :group, column_name: "members_count"
  # If you'd like to also touch the group when `members_count` is updated
  # counter_culture :group, column_name: "members_count", touch: true
end

Now, the Group model will have an up to date count of its members in the members_count column

Multi-level counter-cache

class Product < ActiveRecord::Base
  belongs_to :sub_category
  counter_culture [:sub_category, :category]
end

class SubCategory < ActiveRecord::Base
  has_many :products
  belongs_to :category
end

class Category < ActiveRecord::Base
  has_many :sub_categories
end

Now, the Category model will keep an up-to-date counter-cache in the products_count column of the categories table. This will work with any number of levels.

If you want to have a counter-cache for each level of your hierarchy, then you must add a separate counter cache for each level. In the above example, if you wanted a count of products for each category and sub_category you would change the Product class to:

class Product < ActiveRecord::Base
  belongs_to :sub_category
  counter_culture [:sub_category, :category]
  counter_culture [:sub_category]
end

Customizing the column name

class Product < ActiveRecord::Base
  belongs_to :category
  counter_culture :category, column_name: "products_counter_cache"
end

class Category < ActiveRecord::Base
  has_many :products
end

Now, the Category model will keep an up-to-date counter-cache in the products_counter_cache column of the categories table. This will also work with multi-level counter caches.

Dynamic column name

class Product < ActiveRecord::Base
  belongs_to :category
  counter_culture :category, column_name: proc {|model| "#{model.product_type}_count" }
  # attribute product_type may be one of ['awesome', 'sucky']
end

class Category < ActiveRecord::Base
  has_many :products
end

Delta Magnitude

class Product < ActiveRecord::Base
  belongs_to :category
  counter_culture :category, column_name: :weight, delta_magnitude: proc {|model| model.product_type == 'awesome' ? 2 : 1 }
end

class Category < ActiveRecord::Base
  has_many :products
end

Now the Category model will keep the weight column up to date: awesome products will affect it by a magnitude of 2, others by a magnitude of 1.

You can also use a static multiplier as the delta_magnitude:

class Product < ActiveRecord::Base
  belongs_to :category
  counter_culture :category, column_name: :weight, delta_magnitude: 3
end

class Category < ActiveRecord::Base
  has_many :products
end

Now adding a Product will increase the weight column in its Category by 3; deleting it will decrease it by 3.

Conditional counter cache

class Product < ActiveRecord::Base
  belongs_to :category
  counter_culture :category, column_name: proc {|model| model.special? ? 'special_count' : nil }
end

class Category < ActiveRecord::Base
  has_many :products
end

Now, the Category model will keep the counter cache in special_count up-to-date. Only products where special? returns true will affect the special_count.

If you would like to use this with counter_culture_fix_counts, make sure to also provide the column_names configuration.

Temporarily skipping counter cache updates

If you would like to temporarily pause counter_culture, for example in a backfill script, you can do so as follows:

Review.skip_counter_culture_updates do
  user.reviews.create!
end

user.reviews_count # => unchanged

Totaling instead of counting

Instead of keeping a running count, you may want to automatically track a running total. In that case, the target counter will change by the value in the totaled field instead of changing by exactly 1 each time. Use the :delta_column option to specify that the counter should change by the value of a specific field in the counted object. For example, suppose the Product model table has a field named weight_ounces, and you want to keep a running total of the weight for all the products in the Category model's product_weight_ounces field:

class Product < ActiveRecord::Base
  belongs_to :category
  counter_culture :category, column_name: 'product_weight_ounces', delta_column: 'weight_ounces'
end

class Category < ActiveRecord::Base
  has_many :products
end

Now, the Category model will keep the counter cache in product_weight_ounces up-to-date. The value in the counter cache will be the sum of the weight_ounces values in each of the associated Product records.

The :delta_column option supports all numeric column types, not just :integer. Specifically, :float is supported and tested.

Dynamically over-writing affected foreign keys

class Product < ActiveRecord::Base
  belongs_to :category
  counter_culture :category, foreign_key_values:
      proc {|category_id| [category_id, Category.find_by_id(category_id).try(:parent_category).try(:id)] }
end

class Category < ActiveRecord::Base
  belongs_to :parent_category, class_name: 'Category', foreign_key: 'parent_id'
  has_many :children, class_name: 'Category', foreign_key: 'parent_id'

  has_many :products
end

Now, the Category model will keep an up-to-date counter-cache in the products_count column of the categories table. Each product will affect the counts of both its immediate category and that category's parent. This will work with any number of levels.

Updating timestamps when counts change

By default, counter_culture does not update the timestamp of models when it updates their counter caches. If you would like every change in the counter cache column to result in an updated timestamp, simply set the touch option to true:

  counter_culture :category, touch: true

This is useful when you require your caches to get invalidated when the counter cache changes.

Custom timestamp column

You may also specify a custom timestamp column that gets updated only when a particular counter cache changes:

  counter_culture :category, touch: 'category_count_changed'

With this option, any time the category_counter_cache changes both the category_count_changed and updated_at columns will get updated.

Avoiding deadlocks / executing counter cache updates after commit

Some applications run into issues with deadlocks involving counter cache updates when using this gem. See #263 for information and helpful links on how to avoid this issue.

Another option is to simply defer the update of counter caches to outside of the transaction. This gives up transactional guarantees for your counter cache updates but should resolve any deadlocks you experience. This behavior is disabled by default, enable it on each affected counter cache as follows:

  counter_culture :category, execute_after_commit: true

[NOTE] You need to manually specify the after_commit_action as dependency in the Gemfile to use this feature

...
gem "after_commit_action"
...

You can also pass a Proc for dynamic control. This is useful for temporarily moving the counter cache update inside of the transaction:

  counter_culture :category, execute_after_commit: proc { !Thread.current[:update_counter_cache_in_transaction] }

Aggregating multiple updates into single SQL queries

NOTE: This does not have an effect on Papertrail callbacks

By default, every create/update/destroy will trigger separate UPDATE SQLs for each action and targeted column. For example, if the creation of 1 record needs to increment 3 counter cache columns, a transaction that creates 3 records of this kind will generate 9 UPDATE SQL queries adding +1 to the value of said columns (3 created records x 3 counter cache columns to be incremented by each).

To avoid this, you can wrap the logic that creates/updates/destroys multiple records in CounterCulture.aggregate_counter_updates. This will sum all updates for counter cache columns and will take the latest value for timestamp updates. As a result, you will get just one UPDATE SQL query per target record that updates all of its columns at once.

Execute aggregated SQL queries before transaction COMMIT

ActiveRecord::Base.transaction do
  CounterCulture.aggregate_counter_updates do
    # list of updates
  end # => executes aggregated SQLs
end

Execute aggregated SQL queries after transaction COMMIT

CounterCulture.aggregate_counter_updates do
  ActiveRecord::Base.transaction do
    # list of updates
  end
end # => executes aggregated SQLs

Examples:

Sums multiple counter updates for single column of a single target record
-- Before
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) + 1 WHERE `authors`.`id` = 1
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) + 1 WHERE `authors`.`id` = 1
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) + 1 WHERE `authors`.`id` = 1
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) - 1 WHERE `authors`.`id` = 1

-- After
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) + 2 WHERE `authors`.`id` = 1
-- Before
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) - 1 WHERE `authors`.`id` = 1
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) - 1 WHERE `authors`.`id` = 1

-- After
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) + -2 WHERE `authors`.`id` = 1
-- Before
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) + 1 WHERE `authors`.`id` = 1
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) - 1 WHERE `authors`.`id` = 1

-- After
No query
Combines multiple updates for multiple columns of a single target record
-- Before
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) + 1, updated_at = '2024-06-06 01:00:00' WHERE `authors`.`id` = 1
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) + 1, updated_at = '2024-06-06 02:00:00' WHERE `authors`.`id` = 1

-- After
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) + 2, updated_at = '2024-06-06 02:00:00' WHERE `authors`.`id` = 1
-- Before
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) + 1 WHERE `authors`.`id` = 1
UPDATE `authors` SET `authors`.`works_count` = COALESCE(`authors`.`works_count`, 0) + 1 WHERE `authors`.`id` = 1

-- After
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) + 1, `authors`.`works_count` = COALESCE(`authors`.`works_count`, 0) + 1 WHERE `authors`.`id` = 1
Combines multiple updates of multiple target records
-- Before
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) - 1 WHERE `authors`.`id` = 1
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) - 1 WHERE `authors`.`id` = 1
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) - 1 WHERE `authors`.`id` = 1
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) + 1 WHERE `authors`.`id` = 2
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) + 1 WHERE `authors`.`id` = 2
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) + 1 WHERE `authors`.`id` = 2

-- After
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) + -3 WHERE `authors`.`id` = 1
UPDATE `authors` SET `authors`.`books_count` = COALESCE(`authors`.`books_count`, 0) + 3 WHERE `authors`.`id` = 2

Manually populating counter cache values

You will sometimes want to populate counter-cache values from primary data. This is required when adding counter-caches to existing data. It is also recommended to run this regularly (at BestVendor, we run it once a week) to catch any incorrect values in the counter caches.

Product.counter_culture_fix_counts
# will automatically fix counts for all counter caches defined on Product

Product.counter_culture_fix_counts exclude: :category
# will automatically fix counts for all counter caches defined on Product, except for the :category relation

Product.counter_culture_fix_counts only: :category
# will automatically fix counts only on the :category relation on Product

# :exclude and :only also accept arrays of one level relations
# if you want to fix counts on a more than one level relation you need to use convention below:

Product.counter_culture_fix_counts only: [[:subcategory, :category]]
# will automatically fix counts only on the two-level [:subcategory, :category] relation on Product

Product.counter_culture_fix_counts column_name: :reviews_count
# will automatically fix counts only on the :reviews_count column on Product
# This allows us to skip the columns that have already been processed. This is useful after running big DB changes that affect only one counter cache column.

# :except and :only also accept arrays

Product.counter_culture_fix_counts verbose: true
# prints some logs to STDOUT

Product.counter_culture_fix_counts only: :category, where: { categories: { id: 1 } }
# will automatically fix counts only on the :category with id 1 relation on Product

The counter_culture_fix_counts counts method uses batch processing of records to keep the memory consumption low. The default batch size is 1000 but is configurable like so

# In an initializer
CounterCulture.config.batch_size = 100

or by passing the :batch_size option to the method call

Product.counter_culture_fix_counts batch_size: 100

counter_culture_fix_counts returns an array of hashes of all incorrect values for debugging purposes. The hashes have the following format:

{ entity: which model the count was fixed on,
  id: the id of the model that had the incorrect count,
  what: which column contained the incorrect count,
  wrong: the previously saved, incorrect count,
  right: the newly fixed, correct count }

counter_culture_fix_counts is optimized to minimize the number of queries and runs very quickly.

Similarly to counter_culture, it is possible to update the records' timestamps, when fixing counts. If you would like to update the default timestamp field, pass touch: true option:

Product.counter_culture_fix_counts touch: true

If you have specified a custom timestamps column, pass its name as the value for the touch option:

Product.counter_culture_fix_counts touch: 'category_count_changed'

Parallelizing fix counter cache in multiple workers

The options start and finish are especially useful if you want multiple workers dealing with the same processing queue. You can make worker 1 handle all the records between id 1 and 9999 and worker 2 handle from 10000 and beyond by setting the :start and :finish option on each worker.

! NOTE: the IDs we pass as start and finish here are in fact Category IDs, not Product!

Product.counter_culture_fix_counts start: 10_000
# will fix counts for all counter caches defined on Product from record 10000 and onwards.

Product.counter_culture_fix_counts finish: 10_000
# let's process until 10000 records.

Product.counter_culture_fix_counts start: 1000, finish: 2000
# In worker 1, lets process from 1000 to 2000

Product.counter_culture_fix_counts start: 2001, finish: 3000
# In worker 2, lets process from 2001 to 3000

Fix counter cache using a replica database

When fixing counter caches the number of reads usually vastly exceeds the number of writes. It can make sense to offload the read load to a replica database in this case. Rails 6 introduced native handling of multiple database connections. You can use this to send read traffic to a read-only replica using the option db_connection_builder:

Product.counter_culture_fix_counts db_connection_builder: proc{|reading, block|
  if reading # Count calls will request a reading connection
    Product.connected_to(role: :reading, &block)
  else # Update all calls will request a non-reading connection
    Product.connected_to(role: :writing, &block)
  end
}

Handling dynamic column names

Manually populating counter caches with dynamic column names requires additional configuration:

class Product < ActiveRecord::Base
  belongs_to :category
  counter_culture :category,
      column_name: proc {|model| "#{model.product_type}_count" },
      column_names: {
          ["products.product_type = ?", 'awesome'] => 'awesome_count',
          ["products.product_type = ?", 'sucky'] => 'sucky_count'
      }
  # attribute product_type may be one of ['awesome', 'sucky']
end

You can specify a scope instead of a where condition string for column_names. We recommend providing a Proc that returns a hash instead of directly providing a hash: If you were to directly provide a scope this would load your schema cache on startup which will break things like rake db:migrate.

class Product < ActiveRecord::Base
  belongs_to :category
  scope :awesomes, ->{ where "products.product_type = ?", 'awesome' }
  scope :suckys, ->{ where "products.product_type = ?", 'sucky' }

  counter_culture :category,
      column_name: proc {|model| "#{model.product_type}_count" },
      column_names: -> { {
          Product.awesomes => :awesome_count,
          Product.suckys => :sucky_count
      } }
end

If you would like to avoid this configuration and simply skip counter caches with dynamic column names, while still fixing those counters on the model that are not dynamic, you can pass skip_unsupported:

Product.counter_culture_fix_counts skip_unsupported: true

You can also use context within the block that was provided with the column_names method:

class Product < ActiveRecord::Base
  belongs_to :category
  scope :awesomes, -> (ids) { where(ids: ids, product_type: 'awesome') }

  counter_culture :category,
      column_name: 'awesome_count'
      column_names: -> (context) {
        { Product.awesomes(context[:ids]) => :awesome_count }
      }
end

Product.counter_culture_fix_counts(context: { ids: [1, 2] })

Handling over-written, dynamic foreign keys

Manually populating counter caches with dynamically over-written foreign keys (:foreign_key_values option) is not supported. You will have to write code to handle this case yourself.

Soft-deletes with paranoia or discard

This gem will keep counters correctly updated in Rails 4.2 or later when using paranoia or discard for soft-delete support. However, to ensure that counts are incremented after a restore you have to make sure to set up soft deletion (via acts_as_paranoid or include Discard::Model) before the call to counter_culture in your model:

Paranoia

class SoftDelete < ActiveRecord::Base
  acts_as_paranoid

  belongs_to :company
  counter_culture :company
end

Discard

class SoftDelete < ActiveRecord::Base
  include Discard::Model

  belongs_to :company
  counter_culture :company
end

PaperTrail integration

If you are using the paper_trail gem and would like new versions to be created when the counter cache columns are changed by counter_culture, you can set the with_papertrail option:

class Review < ActiveRecord::Base
  counter_culture :product, with_papertrail: true
end

class Product < ActiveRecord::Base
  has_paper_trail
end

Polymorphic associations

counter_culture now supports polymorphic associations of one level only.

To discover which models need to be updated via counter_culture_fix_counts, counter_culture performs a DISTINCT query on the polymorphic relationship. This query can be expensive so we therefore offer the option (polymorphic_classes) to specify the models' counts that should be corrected:

Image.counter_culture_fix_counts(polymorphic_classes: Product)
# or
Image.counter_culture_fix_counts(polymorphic_classes: [Product, Employee])

Contributing to counter_culture

  • Check out the latest master to make sure the feature hasn't been implemented or the bug hasn't been fixed yet.
  • Check out the issue tracker to make sure someone already hasn't requested it and/or contributed it.
  • Fork the project.
  • Start a feature/bugfix branch.
  • Commit and push until you are happy with your contribution.
  • Make sure to add tests for it. This is important so I don't break it in a future version unintentionally.
  • Please try not to mess with the Rakefile, version, or history. If you want to have your own version, or is otherwise necessary, that is fine, but please isolate to its own commit so I can cherry-pick around it.

Copyright

Copyright (c) 2012-2021 BestVendor, Magnus von Koeller. See LICENSE.txt for further details.